Adaptive memory

From Wikipedia, the free encyclopedia

https://en.wikipedia.org/wiki/Adaptive_memory 

 

An African grasslands scene, similar to the environment experienced by the evolutionary ancestors of modern humans.

Adaptive memory is the study of memory systems that have evolved to help retain survival- and fitness-related information, i.e., that are geared toward helping an organism enhance its reproductive fitness and chances of surviving. One key element of adaptive memory research is the notion that memory evolved to help survival by better retaining information that is fitness-relevant. One of the foundations of this method of studying memory is the relatively little adaptive value of a memory system that evolved merely to remember past events. Memory systems, it is argued, must use the past in some service of the present or the planning of the future. Another assumption under this model is that the evolved memory mechanisms are likely to be domain-specific, or sensitive to certain types of information.

History of adaptive memory research

A recent development in the field of evolutionary psychology, adaptive memory was first proposed in 2007 by James S. Nairne, Sarah R. Thompson, and Josefa N.S. Pandeirada. Evolutionary psychologists often state that humans possess a "stone-age" brain. Therefore, optimal cognitive performance may be found in problems faced by our ancestors, or related to the environments in which our evolutionary ancestors lived. Based on this finding, Nairne and his colleagues proposed that human memory systems are 'tuned' to information relevant to survival.

The first study on the subject of adaptive memory was published in 2007 and its methodology has been replicated many times since. Participants were told to imagine themselves in one of three randomly assigned scenarios. In the Survival condition, they had to imagine being stranded in a grassland area of a foreign land, and needing to find a steady supply of food and water, and protecting themselves from predators. The second condition was the Moving condition. Participants were instructed to imagine themselves moving to a foreign land, needing to locate a new home and transporting their possessions. Finally, the Pleasantness condition asked participants to simply rate the pleasantness of a list of words. In the Survival and Moving conditions, participants were asked to rate the relevance of each word on a list to their imagined situation. Participants were then subject to a surprise recall test. Nairne et al. found what they called the survival advantage.

The survival advantage means that information that is more salient, or relevant, to survival in an ancestral environment has a higher rate of retention than control conditions. This became clear following free recall experiments conducted by Nairne and colleagues. Processing information relevant to survival leads to more information being remembered than most known encoding techniques. One explanation for the survival advantage's improved retention is that this kind of processing taps into a sort of memory 'module' specialized for remembering and processing information that is important for survival. Another proposed explanation is that it leads to more arousal and emotional processing. Because many survival situations are emotionally arousing, retention is enhanced.

Studying survival-based processing

Methodology

An important first step toward a functional analysis of survival processing is thinking about the kinds of problems memory would have evolved to solve. Figuring out selection pressures can be difficult. Post-hoc accounts, also known as just-so stories, are an important problem to avoid. Nairne et al.'s work stressed the importance of a priori predictions, and designing empirical tests. Like other scientific research, a very important element of studying adaptive memory is to consider its basis on empirical evidence and study methodology.

The methodology for testing adaptive memory and the survival advantage in human participants has thus far mostly consisted of ranking lists of words by their relevance to a survival setting (and along control dimensions as well), followed by a recall session. The basic research methodology involves having participants rate a series of words by their fitness relevance. In the control condition, the words were rated for their relevance to moving for a foreign land, and their pleasantness. A surprise recall test is administered, and recall of the listed words in all three conditions is recorded and analyzed.

Findings

General findings

One major finding is that survival processing has been shown to yield better retention than imagery, self-reference and pleasantness, which are all considered to be among the best conditions for remembering learned information. Otgaar, Smeets and van Bergen hypothesized that since visual processing developed earlier than language, there ought to be a survival advantage for images as well as words, and they found such an advantage. Grasslands survival scenarios showed higher retention than near-identical scenarios in which the word 'grasslands' was replaced with 'city' and the word 'predators' replaced with 'attackers'. It is suspected that this result is due to the human mind being attuned to scenarios relevant to our species' ancestral past, even though threats present in a modern urban society are far more relevant today. There was a greater memory recall in both the ancestral and modern survival conditions when compared to pleasantness control conditions, but only the ancestral condition presented significantly greater word recall. Both conditions are fitness-relevant, but there was no memory enhancement for survival processing in the modern context. Additionally, as females typically performed gathering tasks over the span of human evolution, and males performed hunting tasks, research into this gender dichotomy was conducted. No significant data were found. The implications of this research lie in helping to understand how the mind evolved and how it works. The idea that we are able to retain information most relevant to our own survival provides a foundation of research for empirical studying of memory through an evolutionary lens. Understanding the circumstances when memory is at its best can help with the study of the functions of memory as a whole, and help understand what memory is capable of.

True and false memory

The survival processing advantage has been shown to increase both true and false memory in adults and children. True memory refers to the correct retention of information. False memory means remembering something that was never present. A false memory is not necessarily maladaptive. Misremembering can have advantages in certain situations (for example, misremembering an environment with predator tracks as the actual presence of a predator may lead to avoidance of that area in the future). False memory can be seen as a side-effect to an otherwise highly adaptive process.

Planning future acts

One of the most important functions of memory is the ability to use learned information to make predictions in future planning. Adaptive memory states that memory was created and developed by the process of natural selection, so many components of memory systems were important for long-term planning, the importance of which is central to ensuring survival and the passing of genes.

Gathering-related navigation

There exists evidence that the human memory system has evolved to be equipped with a gathering-related navigation system, helping remember the location of gatherable food sources in spatial memory. This was suggested in a recent study that tested spatial memory for various food items, with additional predictions later extended and validated. In the first study, both males and females were shown to have better spatial memory for more caloric foods at a farmer's market than foods with a lower calorie content. It has been suggested that calorie-dense foods are an important resource to be able to gather in an ancestral environment, and the inability to locate this kind of food would put an organism at a disadvantage.

Neurobiological basis

Location of the hippocampus in the brain, an important structure in the function of memory.

Although there has not been any research done on the direct neurological processes that go on during an adaptive memory consolidation, there is a growing body of evidence that the neurotransmitter dopamine modulates the hippocampus, a cortical structure (brain structure) crucial to memory. The release of dopamine has been known to be associated with events of a motivationally important nature, and has a role in the creation of episodic memories and the consolidation thereof. Episodic memories are crucial in the development and implementation of adaptive future behaviours, for which adaptive memory is a very central construct. Functional imaging research conducted during an adaptive memory experiment might be able to provide some insight into the exact brain activity responsible for the phenomenon.

Proximate mechanisms responsible for survival processing

It has been suggested that the recall advantage for survival processing can be attributed to the use of both item-specific processing (the encoding of individual characteristics of items) and relational processing (encoding the relationships between items). It was found that the survival advantage was present when words were encoded using only item-specific or only relational processing, but was eliminated in cases where both methods of encoding were encouraged. Employing both item-specific and relational processing simultaneously makes survival processing unique; the only other phenomenon that also activates both item-specific and relational is self-referential processing. Self-referential processing is one of the proposed explanations for the survival recall advantage. Evidence suggests that it is possible that the survival task might generate self-referential processing, which may be responsible for the reported recall improvements, since considering one's own survival is very self-relevant.

Evolutionary adaptations of memory in non-human species

Black-capped chickadees (Poecile atricapilla) from Alaska showed stronger recall of the locations of cached food than did those from Colorado.

Adaptive memory behaviours have been observed in animal species, as well as humans. Dr. John Garcia discovered that taste aversion conditioning in rats results in a lasting association between sickness and an ingested substance, and that aversion can be established after only one trial. This rapid learning led others to the idea that rats have biological dispositions for learning to associate sickness with a taste memory as a result of its evolutionary history. The results of a study that compared rats and quail in the acquisition of taste aversion suggested that rats rely on their memory of taste to avoid nausea while quail relied on their visual memory. Since the rat is a nocturnal feeder and has poor vision, it was suggested that they rely on taste cues to learn to avoid toxic substances, leading to a highly developed chemical sense system.

In a study of black-capped chickadees (Poecile atricapilla) from Alaska and Colorado, those birds from Alaska cached more food, remembered the hiding places more quickly and readily, and had significantly larger hippocampal volumes than black-capped chickadees from Colorado. This supports what is known as the adaptive specialization hypothesis. The adaptive specialization hypothesis states that animals that hide food should have bigger hippocampi than animals that do not hide food. Additionally, those animals that hide food performed better than non-hiding animals on spatial memory tests. This finding suggests that since the climate tends to be harsh in Alaska, there may be an increased need to remember where food is hidden, to help them survive through the winter.

Polygamous meadow voles have shown to perform better in route and place learning tasks than the monogamous pine vole, likely attributable to their nomadic nature.

Under the pressure of natural selection, sex differences in spatial memory skills can develop under distinct mating systems. Comparing different vole species that are either polygamous or monogamous, researchers found that sex differences in spatial skills are only found in the polygamous mating meadow vole. Further examination found that unlike monogamous pine voles (who travel together on a permanent basis with their partners), polygamous males have increased mobility (compared to the females) in order to reap reproductive benefits. Male meadow voles with good spatial skills are more likely to be able to survive and reproduce. It is suggested that such evolutionary pressure led to greater spatial skills in comparison to female meadow voles and monogamous male pine voles. This suggestion is based on better performance by male meadow voles on route and place learning tasks in laboratory tests. Polygamous voles have a significantly larger relative hippocampus than females of the same species, and no such difference was found between the pine vole sexes. These findings suggest that when better spatial memory is more evolutionarily adaptive, as in polygamous species, these organisms may develop a larger hippocampus, to allow greater spatial memory.

Alternative viewpoints/explanations

Congruity effect

In contrast to Nairne's findings of a survival advantage, it has been found that the superiority of survival processing can be explained by a congruity effect. The congruity effect means that people tend to be better at remembering items if they are congruent with the way they are processed. Congruence between the processing task and target words leads to deeper and more elaborate encoding, which is thought to explain the survival advantage. Butler and colleagues conducted three experiments to test this view; the first replicated Nairne's work and confirmed a survival recall advantage. The second and third experiments controlled for the congruity effect, which was thought to be an underlying explanation for such a memory advantage. A survival scenario along with a robbery scenario were used to rate words instead of the original survival, moving or pleasantness categories. Butler found that, when the materials are controlled for with respect to congruence between type of processing (survival vs robbery scenario) and nature of materials (word lists), survival processing does not always produce superior recall; survival processing did not produce a recall advantage within the congruent conditions where congruity was controlled for.

False memory illusions and net accuracy

It has been proposed that survival processing and recall of survival-relevant materials increases both true and false memory recall. Howe and Derbish suggested that if human memory does benefit from survival processing, this benefit must include both an increase in true recollection of information actually present and also a reduced susceptibility to false memory illusions. Further, Howe and Derbish state, if survival information is more distinctive and processed at an item-specific contextual level, false memory rates should be low. To test this, Nairne's survival/pleasantness experiment was modified to manipulate the type of processing along with the information being processed (eliminated the possibility of recollection of survival-related material due to arousal or emotionality; neutral, negative or survival related). It was concluded that participants in the survival condition had higher rates of false recognition than participants in the pleasantness condition. Evidence for a survival recall advantage has been found to be accompanied by higher false recall rates and results in lower rates of net accuracy (ratio of true recall to recall as a whole). The survival recall advantage appears to occur when only true recall is considered. When total output is taken into account and net accuracy is calculated, the survival recall advantage disappears.

Basic memory processes

Basic memory processes have also been examined in terms of their relation to survival processing, in hopes of explaining the survival recall advantage. Weinstein, Bugg and Roediger contrasted two basic memory processes: schematic processing (the memory performance is made easier and more efficient with the creation of schemas) and self-referential processing (elaboration becomes easier when relating the concept to oneself). Weinstein and colleagues conducted two experiments, the first duplicating Nairne's findings, and the second comparing the survival advantage to schematic and self-referential processing. Weinstein's findings supported Nairne's survival advantage and found it unlikely that the survival advantage can be explained in terms of schematic processing or self-reference.

Falsified assumptions

Adaptive memory research depends on different auxiliary assumptions drawn from evolutionary psychology. One assumption is that memory will be specially tuned to remember information that is processed in a context similar to the environmental context where the adaptation took place, i.e., the African savannah or "grasslands of a foreign land". However, information that was processed for its relevance to an environment involving zombies was remembered better than information processed under the African savannah context and the African context condition did not differ from outer space processing. These experiments present a problem for the assumption of the importance of the ancestral environment because retention was better for information processed under a non-African savannah context, something not predicted by the theory of adaptive memory.

A second assumption of adaptive memory research is that memory evolved because it was beneficial for survival, thus, based on a functional approach, memory should be more sensitive and retention better for information that is processed for its fitness-relevance. The assumption of fitness-relevance has been experimentally evaluated using a number of specific evolutionarily relevant scenarios designed to tap different adaptive mechanisms (or psychological adaptations), for example, a mating mechanism, fear and phobia mechanism, cheater detection mechanism, etc. When words were processed with respect to these different adaptive mechanisms, there was not a processing benefit for these scenarios. This finding suggests that fitness-relevance is too amorphous of a construct to explain the mnemonic benefit found with survival processing and Sandry et al. suggest that research efforts should be directed at identifying the underlying mechanisms and developing a taxonomy of adaptive memory, similar to evolutionary biology.

Future research

As the concept of adaptive memory is a relatively recent development in memory research, there is still much research to be done in this field. One important consideration for further research in this field is adopting of a functional perspective of memory, leading to the important goal of further empirical findings and the refinement of those results obtained thus far. Another important research goal is identifying the precise conditions under which the survival advantage is in effect and those under which it is not. A third critical research focus is determining the specific functional mechanism or mechanisms responsible for this effect. Future research should be conducted with a wide variety of items such as pictures, categorized lists, and content specific materials (for example, those related to food, reproduction, predators and other survival-relevant domains). Finally, neuroimaging research has yet to be done to address any neurological activity that may be different in adaptive memory processing compared to normal conditions. Further research into adaptive memory would be very helpful in understanding more about how exactly the brain works in a survival situation.

Vehicular automation

From Wikipedia, the free encyclopedia
https://en.wikipedia.org/wiki/Vehicular_automation

The ESA Seeker autonomous rover during tests at Paranal

Vehicular automation involves the use of mechatronics, artificial intelligence, and multi-agent systems to assist the operator of a vehicle (car, aircraft, watercraft, or otherwise). These features and the vehicles employing them may be labeled as intelligent or smart. A vehicle using automation for difficult tasks, especially navigation, to ease but not entirely replace human input, may be referred to as semi-autonomous, whereas a vehicle relying solely on automation is called robotic or autonomous. Both of these types are instantiated in today's various self-driving cars, advanced airliner autopilots, drone aircraft, and planetary rovers, as well as guided rockets and missiles. After the invention of the integrated circuit, the sophistication of automation technology increased. Manufacturers and researchers subsequently added a variety of automated functions to automobiles and other vehicles. The technology involved in implementing autonomous vehicles is very expansive, ranging from technological improvements in the vehicle itself to the environment and objects around the vehicle. As the use of automated vehicles increases, they are becoming more influential in human lives. Although automated vehicles bring various benefits, they also come with various concerns. Also, there are still technological challenges that need to be overcome in order to make vehicular automation robust and scalable.

Overview

Automated vehicle system technology hierarchy

An automated driving system is generally an integrated package of individual automated systems operating in concert. Automated driving implies that the driver has delegated the ability to drive (i.e., all appropriate monitoring, agency, and action functions) to the vehicle automation system. Even though the driver may be alert and ready to take action at any moment, the automation system controls all functions.

Automated driving systems are often conditional, meaning that the automation system is capable of automated driving, but not for all conditions encountered in the course of normal operation. Therefore, a human driver is functionally required to initiate the automated driving system, and may or may not do so when driving conditions are within the capability of the system. When the vehicle automation system has assumed all driving functions, the human is no longer driving the vehicle but continues to assume responsibility for the vehicle's performance as the vehicle operator. The automated vehicle operator is not functionally required to actively monitor the vehicle's performance while the automation system is engaged, but the operator must be available to resume driving within several seconds of being prompted to do so, as the system has limited conditions of automation. While the automated driving system is engaged, certain conditions may prevent real-time human input, but for no more than a few seconds. The operator is able to resume driving at any time subject to this short delay. When the operator has resumed all driving functions, he or she is the vehicle's driver again.

Fully autonomous cars and trucks that drive us instead of us driving them will become a reality. These self-driving vehicles ultimately will integrate onto U.S. roadways by progressing through six levels of driver assistance technology advancements in the coming years. This includes everything from no automation (where a fully engaged driver is required at all times), to full autonomy (where an automated vehicle operates independently, without a human driver).

An automated driving system is defined in an proposed amendment to Article 1 of the Vienna Convention on Road Traffic:

(ab) "Automated driving system" refers to a vehicle system that uses both hardware and software to exercise dynamic control of a vehicle on a sustained basis.

(ac) "Dynamic control" refers to carrying out all the real-time operational and tactical functions required to move the vehicle. This includes controlling the vehicle’s lateral and longitudinal motion, monitoring the road environment, responding to events in the road traffic environment, and planning and signalling for manoeuvres.

This amendment will enter into force on 14 July 2022, unless it is rejected before 13 January 2022.

An automated driving feature must be described sufficiently clearly so that it is distinguished from an assisted driving feature.

— SMMT

There are two clear states – a vehicle is either assisted with a driver being supported by technology or automated where the technology is effectively and safely replacing the driver.

— SMMT

Autonomy levels

Main article: Self-driving_car § Classification

Further information: Automatic train operation § Types of train automation

Autonomy in vehicles is often categorized in six levels: The level system was developed by the Society of Automotive Engineers (SAE).

• Level 0: No automation.
• Level 1: Driver assistance - The vehicle can control either steering or speed autonomously in specific circumstances to assist the driver.
• Level 2: Partial automation - The vehicle can control both steering and speed autonomously in specific circumstances to assist the driver.
• Level 3: Conditional automation - The vehicle can control both steering and speed autonomously under normal environmental conditions, but requires driver oversight.
• Level 4: High automation - The vehicle can complete travel autonomously under normal environmental conditions, not requiring driver oversight.
• Level 5: Full autonomy - The vehicle can complete travel autonomously in any environmental conditions.

Level 0 refers, for instance, to vehicles which do not have adaptive cruise control.

Level 1 and 2 refer to vehicles where one part of the driving task is performed by the vehicle advanced driver-assistance systems (ADAS) under the responsibility/accountability/liability of the driver.

From level 3, the driver can conditionally transfer the driving task to the vehicle, but the driver must take back control when the conditional automation is no longer available. For instance an automated traffic jam pilot can drive in the traffic jam but the driver should take back control when traffic jam is over.

Level 5 refers to a vehicle which does not need any (human) driver.

"Level 2+" or "semi-automated" is a kind of enhanced level 2 where some manufacturers are ready to provide more features than the basic features of a level 2 system, but manufacturers and regulators are not yet ready for SAE level 3. This led to the introduction of the informal notion of "enhanced level 2" or "level 2+" or "semi-automated" — a dominant ADAS trend in 2021 — which is a level 2 with additional safety and comfort. For instance, a low cost level 2+ vehicle can include interior-monitoring technologies to ensure driver attention, adaptive merging for when vehicles are entering or exiting the highway, and new kinds of enhanced automatic emergency braking (AEB) for pedestrian, cyclist and motorcyclist safety. Level 2+ can also include lane change and overtaking.

The levels can be roughly understood as Level 0 - no automation; Level 1 - hands on/shared control; Level 2 - hands off; Level 3 - eyes off; Level 4 - mind off, and Level 5 - steering wheel optional.

As of December 2021, level 3 remains a marginal portion of the market, with only one hundred level 3 Honda Legend cars available in the Japanese market. It is possible that level 3 remains a marginal portion of the market until 2025.

Technology used in vehicular automation

The primary means of implementing autonomous vehicles is through the use of Artificial Intelligence (AI). In order for full autonomous vehicles to be implemented, the lower levels of automation must be thoroughly tested and implemented before moving on to the next level. Through implementing autonomous systems, such as navigation, collision avoidance and steering, autonomous vehicle manufacturers work towards higher levels of autonomy by designing and implementing different systems of the car. These autonomous systems, along with the use of artificial intelligence methods, can use the machine learning aspect of AI in order for the vehicle to control each of the other autonomous systems and processes. Thus, autonomous vehicle manufacturers are researching and developing appropriate AI specifically for autonomous vehicles. While many of these companies are continuously developing technologies to be implemented into their autonomous vehicles, the general consensus is that the underlying technology is still in need of further development before fully autonomous vehicles are possible.

Arguably one of the most important systems of any autonomous vehicle, the perception system must be fully developed and well-tested in order for autonomy to advance. With the development and implementation of the perception system on autonomous vehicles, much of the safety standards of autonomous vehicles are being addressed by this system, which places an unequivocal emphasis on it to be flawless, as human lives would be subject to harm if a faulty system were to be developed. The main purpose for the perception system is to constantly scan the surrounding environment and determine which objects in the environment pose a threat to vehicles. In a sense, the perception system's main goal is to act like human perception, allowing the system to sense hazards and to prepare or correct for these hazards. In terms of the detection part of the perception system, many solutions are being tested for accuracy and compatibility, such as radar, lidar, sonar and moving image processing.

With the development of these autonomous subsystems of the car, autonomous vehicle manufacturers have already developed systems which act as assistance features on a vehicle. These systems are known as advanced driver-assistance systems, and contain systems to do such actions as parallel parking and emergency braking. Along these systems, autonomous navigation systems play a role in the development of autonomous vehicles. In implementing the navigation system, there are two ways in which navigation can be implemented: sensing from one vehicle to another or sensing from the infrastructure. These navigation systems would work in tandem with already well established navigation systems, such as the Global Positioning System (GPS), and be able to process route information, detecting such things as traffic jams, tolls and or road construction. From this information, the vehicle can then take the appropriate action to either avoid the area or plan accordingly. However, there may be problems in using this method, such as outdated information, in which case vehicle to infrastructure communication can play a large role in constantly having up-to-date information. An instance of this is having street signs and other regulatory markers display information to the vehicle, which allows the vehicle to make decisions based on the current information.

Along with the development of autonomous vehicles, many of these vehicles are expected to be primarily electric, meaning that the main power source of the vehicle will be electric-based rather than fossil fuel-based. Along with that, there comes the extra demand on autonomous vehicle manufacturers to produce higher quality electric cars in order to implement all the autonomous systems associated with the vehicle. However, much of modern-day vehicle components can still be used in autonomous vehicles, such as the use of the automatic transmissions and operator protection equipment like airbags.

In consideration of the development of autonomous vehicles, companies also are considering operator preferences and needs. These instances include allowing the user to minimize time, follow a precise route and accommodate any possible disabilities that the operator may have. Along with accommodating the driver, autonomous vehicles also impose a technological factor onto the environment around it, generally needing a higher sense of connectivity in the vehicle's environment. With this new factor to consider, many urban governments are considering becoming a smart city in order to provide a sufficient foundation for autonomous vehicles. Along these same lines of the vehicle's environment accommodating the vehicle, the user of these vehicles may also have to be technologically connected in order to operate these autonomous vehicles. With the advent of smartphones, it is predicted that autonomous vehicles will be able to have this connection with the user's smartphone or other technological devices similar to a smartphone.

Success in the technology

AAA Foundation for Traffic Safety conducted a test of two automatic emergency braking systems: those designed to prevent crashes and others that aim to make a crash less severe. The test looked at popular models like the 2016 Volvo XC90, Subaru Legacy, Lincoln MKX, Honda Civic and Volkswagen Passat. Researchers tested how well each system stopped when approaching both a moving and nonmoving target. It found that systems capable of preventing crashes reduced vehicle speeds by twice that of the systems designed to merely mitigate crash severity. When the two test vehicles traveled within 30 mph of each other, even those designed to simply lessen crash severity avoided crashes 60 percent of the time.

The success in the automated driving system has been known to be successful in situations like rural road settings. Rural road settings would be a setting in which there is lower amounts of traffic and lower differentiation between driving abilities and types of drivers. "The greatest challenge in the development of automated functions is still inner-city traffic, where an extremely wide range of road users must be considered from all directions." This technology is progressing to a more reliable way of the automated driving cars to switch from auto-mode to driver mode. Auto-mode is the mode that is set in order for the automated actions to take over, while the driver mode is the mode set in order to have the operator controlling all functions of the car and taking the responsibilities of operating the vehicle (Automated driving system not engaged).

This definition would include vehicle automation systems that may be available in the near term—such as traffic-jam assist, or full-range automated cruise control—if such systems would be designed such that the human operator can reasonably divert attention (monitoring) away from the performance of the vehicle while the automation system is engaged. This definition would also include automated platooning (such as conceptualized by the SARTRE project).

The SARTRE Project

The SARTRE project's main goal is to create platooning, a train of automated cars, that will provide comfort and have the ability for the driver of the vehicle to arrive safely to a destination. Along with the ability to be along the train, drivers that are driving past these platoons, can join in with a simple activation of the automated driving system that correlates with a truck that leads the platoon. The SARTRE project is taking what we know as a train system and mixing it with automated driving technology. This is intended to allow for an easier transportation through cities and ultimately help with traffic flow through heavy automobile traffic.

In some parts of the world the self-driving car has been tested in real life situations such as in Pittsburgh. The Self-driving Uber has been put to the test around the city, driving with different types of drivers as well as different traffic situations. Not only have there been testing and successful parts to the automated car, but there has also been extensive testing in California on automated busses. The lateral control of the automated buses uses magnetic markers such as the platoon at San Diego, while the longitudinal control of the automated truck platoon uses millimeter wave radio and radar. Current examples around today's society include the Google car and Tesla's models. Tesla has redesigned automated driving, they have created car models that allow drivers to put in the destination and let the car take over. These are two modern day examples of the automated driving system cars.

Risks and liabilities

See also: Computer security § Automobiles, and Autonomous car liability

Many automakers such as Ford and Volvo have announced plans to offer fully automated cars in the future. Extensive research and development is being put into automated driving systems, but the biggest problem automakers cannot control is how drivers will use system. Drivers are stressed to stay attentive and safety warnings are implemented to alert the driver when corrective action is needed. Tesla Motor's has one recorded incident that resulted in a fatality involving the automated driving system in the Tesla Model S. The accident report reveals the accident was a result of the driver being inattentive and the autopilot system not recognizing the obstruction ahead.

Another flaw with automated driving systems is that in situations where unpredictable events such as weather or the driving behavior of others may cause fatal accidents due to sensors that monitor the surroundings of the vehicle not being able to provide corrective action.

To overcome some of the challenges for automated driving systems, novel methodologies based on virtual testing, traffic flow simulation and digital prototypes have been proposed, especially when novel algorithms based on Artificial Intelligence approaches are employed which require extensive training and validation data sets.

The implementation of automated driving systems poses the possibility of changing build environments in urban areas, such as the expansion of suburban areas due to the increased ease of mobility.

Challenges

Around 2015, several self-driving car companies including Nissan and Toyota promised self-driving cars by 2020. However, the predictions turned out to be far too optimistic.

There are still many obstacles in developing fully autonomous Level 5 vehicles, which is able to operate in any conditions. Currently, companies are focused on Level 4 automation, which is able to operate under certain environmental circumstances.

There is still debate about what an autonomous vehicle should look like. For example, whether to incorporate lidar to autonomous driving systems is still being argued. Some researchers have come up with algorithms utilizing camera-only data that achieve the performance that rival those of lidar. On the other hand, camera-only data sometimes draw inaccurate bounding boxes, and thus lead to poor predictions. This is due to the nature of superficial information that stereo cameras provide, whereas incorporating lidar gives autonomous vehicles precise distance to each point on the vehicle.

Technical challenges

• Software Integration: Because of the large number of sensors and safety processes required by autonomous vehicles, software integration remains a challenging task. A robust autonomous vehicle should ensure that the integration of hardware and software can recover from component failures.
• Prediction and trust among autonomous vehicles: Fully autonomous cars should be able to anticipate the actions of other cars like humans do. Human drivers are great at predicting other drivers' behaviors, even with a small amount of data such as eye contact or hand gestures. In the first place, the cars should agree on traffic rules, whose turn it is to drive in an intersection, and so on. This scales into a larger issue when there exists both human-operated cars and self-driving cars due to more uncertainties. A robust autonomous vehicle is expected to improve on understanding the environment better to address this issue.
• Scaling up: The coverage of autonomous vehicles testing could not be accurate enough. In cases where heavy traffic and obstruction exist, it requires faster response time or better tracking algorithms from the autonomous vehicles. In cases where unseen objects are encountered, it's important that the algorithms are able to track these objects and avoid collisions.

Societal challenges

One critical step to achieve the implementation of autonomous vehicles is the acceptance by the general public. It is an important ongoing research because it provides guidelines for the automobile industry to improve their design and technology. Studies have shown that many people believe that using autonomous vehicles is safer, which underlines the necessity for the automobile companies to assure that autonomous vehicles improve safety benefits. The TAM research model breaks down important factors that affect the consumer's acceptance into: usefulness, ease to use, trust, and social influence.

• The usefulness factor studies whether or not autonomous vehicles are useful in that they provide benefits that save consumers' time and make their lives simpler. How well the consumers believe autonomous vehicles will be useful compared to other forms of transportation solutions is a determining factor.
• The ease to use factor studies the user-friendliness of the autonomous vehicles. While the notion that consumers care more about ease to use than safety has been challenged, it still remains an important factor that has indirect effects on the public's intention to use autonomous vehicles.
• The trust factor studies the safety, data privacy and security protection of autonomous vehicles. A more trusted system has a positive impact on the consumer's decision to use autonomous vehicles.
• The social influence factor studies whether the influence of others would influence consumer's likelihood of having autonomous vehicles. Studies have shown that the social influence factor is positively related to behavioral intention. This might be due to the fact that cars traditionally serve as a status symbol that represents one's intent to use and his social environment.

Regulatory challenges

Real-time testing of autonomous vehicles is an inevitable part of the process. At the same time, vehicular automation regulators are faced with challenges to protect public safety and yet allow autonomous vehicle companies to test their products. Groups representing autonomous vehicle companies are resisting most regulations, whereas groups representing vulnerable road users and traffic safety are pushing for regulatory barriers. To improve traffic safety, the regulators are encouraged to find a middle ground that protects the public from immature technology while allowing autonomous vehicle companies to test the implementation of their systems. There have also been proposals to adopt the aviation automation safety regulatory knowledge into the discussions of safe implementation of autonomous vehicles, due to the experience that has been gained over the decades by the aviation sector on safety topics.

Ground vehicles

Further information: Unmanned ground vehicles

Ground vehicles employing automation and teleoperation include shipyard gantries, mining trucks, bomb-disposal robots, robotic insects, and driverless tractors.

There are a lot of autonomous and semi-autonomous ground vehicles being made for the purpose of transporting passengers. One such example is the free-ranging on grid (FROG) technology which consists of autonomous vehicles, a magnetic track and a supervisory system. The FROG system is deployed for industrial purposes in factory sites and has been in use since 1999 on the ParkShuttle, a PRT-style public transport system in the city of Capelle aan den IJssel to connect the Rivium business park with the neighboring city of Rotterdam (where the route terminates at the Kralingse Zoom metro station). The system experienced a crash in 2005 that proved to be caused by a human error.

Applications for automation in ground vehicles include the following:

• Vehicle tracking system system ESITrack, Lojack
• Rear-view alarm, to detect obstacles behind.
• Anti-lock braking system (ABS) (also Emergency Braking Assistance (EBA)), often coupled with Electronic brake force distribution (EBD), which prevents the brakes from locking and losing traction while braking. This shortens stopping distances in most cases and, more importantly, allows the driver to steer the vehicle while braking.
• Traction control system (TCS) actuates brakes or reduces throttle to restore traction if driven wheels begin to spin.
• Four wheel drive (AWD) with a centre differential. Distributing power to all four wheels lessens the chances of wheel spin. It also suffers less from oversteer and understeer.
• Electronic Stability Control (ESC) (also known for Mercedes-Benz proprietary Electronic Stability Program (ESP), Acceleration Slip Regulation (ASR) and Electronic differential lock (EDL)). Uses various sensors to intervene when the car senses a possible loss of control. The car's control unit can reduce power from the engine and even apply the brakes on individual wheels to prevent the car from understeering or oversteering.
• Dynamic steering response (DSR) corrects the rate of power steering system to adapt it to vehicle's speed and road conditions.

Research is ongoing and prototypes of autonomous ground vehicles exist.

Cars

See also: Self-driving car

Extensive automation for cars focuses on either introducing robotic cars or modifying modern car designs to be semi-autonomous.

Semi-autonomous designs could be implemented sooner as they rely less on technology that is still at the forefront of research. An example is the dual mode monorail. Groups such as RUF (Denmark) and TriTrack (USA) are working on projects consisting of specialized private cars that are driven manually on normal roads but also that dock onto a monorail/guideway along which they are driven autonomously.

As a method of automating cars without extensively modifying the cars as much as a robotic car, Automated highway systems (AHS) aims to construct lanes on highways that would be equipped with, for example, magnets to guide the vehicles. Automation vehicles have auto-brakes named as Auto Vehicles Braking System (AVBS). Highway computers would manage the traffic and direct the cars to avoid crashes.

In 2006, The European Commission has established a smart car development program called the Intelligent Car Flagship Initiative. The goals of that program include:

• Adaptive cruise control
• Lane departure warning system
• Project AWAKE for drowsy drivers

There are plenty of further uses for automation in relation to cars. These include:

• Assured Clear Distance Ahead
• Adaptive headlamps
• Advanced Automatic Collision Notification, such as OnStar
• Intelligent Parking Assist System
• Automatic Parking
• Automotive night vision with pedestrian detection
• Blind spot monitoring
• Driver Monitoring System
• Robotic car or self-driving car which may result in less-stressed "drivers", higher efficiency (the driver can do something else), increased safety and less pollution (e.g. via completely automated fuel control)
• Precrash system
• Safe speed governing
• Traffic sign recognition
• Following another car on a motorway – "enhanced" or "adaptive" cruise control, as used by Ford and Vauxhall
• Distance control assist – as developed by Nissan
• Dead man's switch – there is a move to introduce deadman's braking into automotive application, primarily heavy vehicles, and there may also be a need to add penalty switches to cruise controls.

Singapore also announced a set of provisional national standards on January 31, 2019, to guide the autonomous vehicle industry. The standards, known as Technical Reference 68 (TR68), will promote the safe deployment of fully driverless vehicles in Singapore, according to a joint press release by Enterprise Singapore (ESG), Land Transport Authority (LTA), Standards Development Organisation and Singapore Standards Council (SSC).

Shared autonomous vehicles

See also: Robotaxi

 

Parkshuttle

 

Navya Autonom Shuttle

 

Easymile EZ10

Since 1999 the 12-seat/10-standing ParkShuttle has been operating on an 1.8 kilometres (1.1 mi) exclusive right of way in the city of Capelle aan den IJssel in The Netherlands. The system uses small magnets in the road surface to allow the vehicle to determine its position. The use of shared autonomous vehicles was trialed around 2012 in a hospital car park in Portugal. From 2012 to 2016 the European Union funded CityMobil2 project examined the use of shared autonomous vehicles and passenger experience including short term trials in seven cities. This project led to the development of the EasyMile EZ10.

Following recent developments in autonomous cars, shared autonomous vehicles are now able to run in mixed traffic without the need for embedded guidance markers. So far the focus has been on low speed, 20 miles per hour (32 km/h), with short, fixed routes for the "last mile" of journeys. This means issues of collision avoidance and safety are significantly less challenging than those for automated cars, which seek to match the performance of conventional vehicles. Many trials have been undertaken, mainly on quiet roads with little traffic or on public pathways or private roadways and specialised test sites. The capacity of different models varies significantly, between 6-seats and 20-seats. (Above this size there are conventional buses that have driverless technology installed.)

In July 2018 Baidu stated it had built 100 of its 8-seat Apolong model, with plans for commercial sales. As of July 2021 they have not gone into volume production.

In August 2020 it was reported there were 25 autonomous shuttle manufacturers, including the 2GetThere, Local Motors, Navya, Baidu, Easymile, Toyota and Ohmio.

Recent prototypes vehicles, showcased in 2020, have top speeds equal to ordinary traffic. These include the six-passenger Cruise "Origin" capable of "highway speeds" and the four-passenger 'Zoox' by Zoox capable up to 75 mph.

In December 2020 Toyota showcased it's 20-passenger "e-Palette" vehicle, which is due to be used at the 2021 Tokyo Olympic Games. Toyota has announced it intends to have the vehicle available for commercial applications before 2025.

In January 2021 Navya released an investor report which predicted global autonomous shuttle sales will reach 12,600 units by 2025, with a market value of EUR 1.7 billion.

In May 2021 Cruise, majority owned by General Motors, announced they expected mass production of the Origin driverless shuttle would commence in 2023. In 2020 the cost per vehicle of the Origin, manufactured at scale, was estimated at $50,000.

In June 2021 Chinese maker Yutong claimed to have delivered 100 models of its 10-seat Xiaoyu 2.0 autonomous bus for use in Zhengzhou. Testing has been carried out in a number of cities since 2019 with trials open to the public due to commence in July 2021.

Planned utilisation

In January 2017 it was announced the ParkShuttle system in the Netherlands will be renewed and expanded including extending the route network beyond the exclusive right of way so vehicles will run in mixed traffic on ordinary roads. The plans were delayed and the extension into mixed traffic is now expected in 2021.

Autonomous shuttles are already in use on some private roads, such as at the Yutong factory in Zhengzhou where they are used to transport workers between buildings of the world's largest bus factory.

In December 2016 the Jacksonville Transportation Authority has announced its intention to replace the Jacksonville Skyway monorail with driverless vehicles that would run on the existing elevated superstructure as well as continue onto ordinary roads. The project has since been named the "Ultimate Urban Circulator" or "U2C" and testing has been carried out on shuttles from six different manufacturers. The cost of the project is estimated at $379 million.

Trials

A large number of trials have been conducted since 2016, with most involving only one vehicle on a short route for a short period of time and with an onboard conductor. The purpose of the trials has been to both provide technical data and to familiarize the public with the driverless technology. A 2021 survey of over 100 shuttle experiments across Europe concluded that low speed - 15–20 kilometres per hour (9.3–12.4 mph) - was the major the barrier to implementation of autonomous shuttle buses. The current cost of the vehicles at €280,000 and the need for onboard attendants were also issues.

In June 2021 it was reported Cruise had begun assembly of 100 pre-production Origin vehicles for validation testing and permits for driverless testing had been issued to them by the State of California.

Company/Location	Details
Navya "Arma" in Neuhausen am Rheinfall	In October 2016, BestMile started trials in Neuhausen am Rheinfall, claiming to be the world's first solution for managing hybrid fleets with both autonomous and non-autonomous vehicles. And the test ended in October 2021.
Local Motors "Olli"	At the end of 2016, the Olli was tested in Washington D.C. In 2020, a four month trial was undertaken at the United Nations ITCILO campus in Turin, Italy to provide transport shuttle to employees and guests within the campus.
Navya "Autonom"	Navya claimed in May 2017 to have carried almost 150,000 passengers across Europe with trials in Sion, Cologne, Doha, Bordeaux and the nuclear power plant at Civaux as well as Las Vegas and Perth. Ongoing public trials are underway in Lyon, Val Thorens and Masdar City. Other trials on private sites are underway at University of Michigan since 2016, at Salford University and the Fukushima Daini Nuclear Power Plant since 2018.
Texas A&M	In August 2017, a driverless four seat shuttle was trialed at Texas A&M university as part of its "Transportation Technology Initiative" in a project run by academics and students on the campus. Another trial, this time using Navya vehicles, was run in 2019 from September to November.
RDM Group "LUTZ Pathfinder"	In October 2017, RDM Group began a trial service with two seat vehicles between Trumpington Park and Ride and Cambridge railway station along the guided busway, for possible use as an after hours service once the regular bus service has stopped each day.
EasyMile "EZ10"	EasyMile has had trials longer term trials at Wageningen University and Lausanne as well as short trials in Darwin, Dubai, Helsinki, San Sebastian, Sophia Antipolis, Bordeaux and Tapei In December 2017 a trial began in Denver running at 5 miles per hour (8.0 km/h) on a dedicated stretch of road. EasyMile was operating in ten U.S. states, including California, Florida, Texas, Ohio, Utah, and Virginia before U.S. service was suspended after a February 2020 injury. In August 2020 EasyMile was operating shuttles in 16 cities across the United States, including Salt Lake City, Columbus, Ohio, and Corpus Christi, Texas. In October 2020 a new trial was launched in Fairfax, Virginia. In August 2021 a one year trial was launched at the Colorado School of Mines in Golden, Colorado. The trial will use nine vehicles (with seven active at any time) and will provide a 5-10 minute service along three routes at a maximum speed of 12 mph (19 km/h). At the time of launch this is the largest such trial in the United States. In November 2021, EasyMile has become the first driverless solutions provider in Europe authorized to operate at Level 4 in mixed traffic, on a public road. "EZ10" has been making test runs on a medical campus in the southwestern city of Toulouse since March.
Westfield Autonomous Vehicles "POD"	In 2017 and 2018, using a modified version of the UltraPRT called "POD" , four vehicles were used as part of the GATEway project trial conducted in Greenwich in south London on a 3.4 kilometres (2.1 mi) route. A number of other trials have been conducted Birmingham, Manchester, Lake District National Park, University of the West of England and Filton Airfield.
Next Future Transportation "pods" in Dubai	In February 2018, the ten passenger (six seated), 12 miles per hour (19 km/h), autonomous pods which are capable of joining to form a bus, were demonstrated at the World Government Summit in Dubai. The demonstration was a collaboration with between Next-Future and Dubai's Roads and Transport Authority and the vehicles are under consideration for deployment there.
"Apolong/Apollo"	In July 2018, a driverless eight seater shuttle bus was trialed at the 2018 Shanghai expo after tests in Xiamen and Chongqing cities as part of Project Apollo, a mass-produced autonomous vehicle project launched by a consortium including Baidu.
Jacksonville Transportation Authority	Since December 2018, the Jacksonville Transportation Authority has been using a 'test and learn' site at the Florida State College at Jacksonville to evaluate vehicles from different vendors as part of its plan for the Ultimate Urban Circulator (U2C). Among the six vehicles tested are the Local Motors "Olli 2.0", Navya "Autonom" and EasyMile "EZ10".
2getthere "ParkShuttle" in Brussels	In 2019, trials were held at Brussels Airport and at Nanyang Technological University in Singapore.
Ohmio "Lift" in Christchurch	In 2019, Trials with their 15-person shuttle were conducted in New Zealand at Christchurch Airport and at the Christchurch Botanic Gardens in 2020.
Yutong "Xioayu"	Testing with the first generation vehicle in 2019 at the Boao Forum for Asia and in Zhengzhou. The 10-seat second generation vehicle has been delivered to Guangzhou, Nanjing, Zhengzhou, Sansha, Changsha with public trials due to commence in July 2021 in Zhengzhou.
ARTC "WinBus" in Changhua city	In July 2020, a trial service began in Changhua city in Taiwan, connecting four tourism factories in Changhua Coastal Industrial Park along a 7.5 km (4.7 mi), with plans to extend the route to 12.6 km (7.8 mi) to serve tourist destinations. In January 2021, Level 4 "WinBus" got a license for one-year experimental sandbox operation.
Yamaha Motor "Land Car" based "ZEN drive Pilot" in Eiheiji town, Fukui prefecture	In December 2020, Eiheiji town started test operation of driverless autonomous driving mobility services by making use of a remotely-operated autonomous driving system. AIST Human-Centered Mobility Research Center (HCMRC) modified Yamaha Motor's electric "Land Car" and the tracing road of an abandoned Eiheiji railway line. This system was legally approved as Level 3.
WeRide "Mini Robobus"	In January 2021, WeRide began testing it's Mini Robobus on Guangzhou International Bio Island. In June 2021, the company also launched trials at Nanjing.
Toyota "e-Palette" in Chūō, Tokyo	During the 2021 Tokyo Summer Olympics, a fleet of 20 vehicles was used to ferry athletes and others around the Athletes' Village. Each vehicle could carry 20 people or 4 wheelchairs and had a top speed of 20 mph (32 km/h). (The event also used 200 driver operated variants called the "Accessible People Movers (APM)", to take athletes to their events.) On August 27, 2021, Toyota suspended all "e-Pallete" services at the Paralympics after a vehicle collided with and injured a visually impaired pedestrian, and restarted on 31 with improved safety measures.
Hino "Poncho Long" tuned by Nippon Mobility in Shinjuku, Tokyo	In November 2021, Tokyo Metropolitan Government starts three trials. As one of the three, a lead contractor Keio Dentetsu Bus is going to overcome unique and difficult conditions in operation in the central of the megalopolis.

Vehicle names are in "quotes"

Motorcycles

Several self-balancing autonomous motorcycles were demonstrated in 2017 and 2018 from BMW, Honda and Yamaha.

Company/Location	Details
Honda motorcycle	Inspired by the Uni-cub, Honda implemented their self-balancing technology into their motorcycles. Due to the weight of motorcycles, it is often a challenge for motorcycle owners to keep balance of their vehicles at low speeds or at a stop. Honda's motorcycle concept has a self-balancing feature that will keep the vehicle upright. It automatically lowers the center of balance by extending the wheelbase. It then takes control of the steering to keep the vehicle balanced. This allows users to navigate the vehicle more easily when walking or driving in stop and go traffic. However, this system is not for high speed driving.
BMWs Motorrad Vision concept motorcycle	BMW Motorrad developed the ConnectRide self driving motorcycle in order to push the boundaries of motorcycle safety. The autonomous features of the motorcycle include emergency braking, negotiating intersections, assisting during tight turns, and front impact avoidance. These are features similar to current technologies that are being developed and implemented in autonomous cars. This motorcycle can also fully drive on its own at normal driving speed, making turns and returning to a designated location. It lacks the self standing feature that Honda has implemented.
Yamaha’s riderless motorcycle	“Motoroid” can hold its balance, autonomous driving, recognizing riders and going to a designated location with a hand gesture. Yamaha utilized the “Human beings react a hell of a lot quicker” research philosophy into the motoroid. The idea is that the autonomous vehicle is not attempting to replace human beings, but to augment the abilities of the human with advanced technology. They have tactile feedback such as a gentle squeeze to a rider's lower back as a reassuring caress at dangerous speeds, as if the vehicle was responding and communicating with the rider. Their goal is to “meld” the machine and human together to form one experience.
Harley-Davidson	While their motorcycles are popular, one of the largest problems of owning a Harley-Davidson is the reliability of the vehicle. It is difficult to manage the weight of the vehicle at low speeds and picking it up from the ground can be a difficult process even with correct techniques. In order to attract more customers, they filed a patent for having a gyroscope at the back of the vehicle that will keep the balance of the motorcycle for the rider at low speeds. After 3 miles per hour, the system disengages. However anything below that, the gyroscope can handle the balance of the vehicle which means it can balance even at a stop. This system can be removed if the rider feels ready without it (meaning it is modular).

Buses

The United Kingdom's first autonomous bus, currently on trial with Stagecoach Manchester

Autonomous buses are proposed as well as self driving cars and trucks. Grade 2 level automated minibuses were trialed for a few weeks in Stockholm. China has also a small fleet of self-driving public buses in the tech district of Shenzhen, Guangdong.

The first autonomous bus trial in the United Kingdom commenced in mid-2019, with an Alexander Dennis Enviro200 MMC single-decker bus modified with autonomous software from Fusion Processing able to operate in driverless mode within Stagecoach Manchester's Sharston bus depot, performing tasks such as driving to the washing station, refuelling point and then parking up at a dedicated parking space in the depot. The first passenger-carrying driverless bus trial in the United Kingdom is expected to commence by 2021, with a fleet of five identical vehicles to the Manchester trial used on a 14 miles (23 km) Stagecoach Fife park-and-ride route across the Forth Road Bridge, from the north bank of the Forth to Edinburgh Park station.

In July 2020 in Japan, AIST Human-Centered Mobility Research Center (HCMRC) with Nippon Koei and Isuzu started a series of demonstration tests for mid-sized buses, Isuzu "Erga Mio" with autonomous driving systems, in five areas; Ōtsu city in Shiga prefecture, Sanda city in Hyōgo Prefecture and other three areas in sequence.

Trucks

Further information: Self-driving truck

The concept for autonomous vehicles has been applied for commercial uses, such as autonomous or nearly autonomous trucks.

Companies such as Suncor Energy, a Canadian energy company, and Rio Tinto Group were among the first to replace human-operated trucks with driverless commercial trucks run by computers. In April 2016, trucks from major manufacturers including Volvo and the Daimler Company completed a week of autonomous driving across Europe, organized by the Dutch, in an effort to get self-driving trucks on the road. With developments in self-driving trucks progressing, U.S. self-driving truck sales is expected to reach 60,000 by 2035 according to a report released by IHS Inc. in June 2016.

As reported in June 1995 in Popular Science Magazine, self-driving trucks were being developed for combat convoys, whereby only the lead truck would be driven by a human and the following trucks would rely on satellite, an inertial guidance system and ground-speed sensors. Caterpillar Inc. made early developments in 2013 with the Robotics Institute at Carnegie Mellon University to improve efficiency and reduce cost at various mining and construction sites.

In Europe, the Safe Road Trains for the Environment is such an approach.

From PWC's Strategy& Report, self driving trucks will be the source of a lot of concern around how this technology will impact around 3 million truck drivers in the US, as well as 4 million employees in support of the trucking economy in gas stations, restaurants, bars and hotels. At the same time, some companies like Starsky, are aiming for Level 3 Autonomy, which would see the driver playing a control role around the truck's environment. The company's project, remote truck driving, would give truck drivers a greater work-life balance, enabling them to avoid long periods away from their home. This would however provoke a potential mismatch between the driver's skills with the technological redefinition of the job.

Companies that buy driverless trucks could massively cut down on costs: human drivers will no longer be required, companies' liabilities due to truck accidents will diminish, and productivity will increase (as the driverless truck doesn't need to rest). The usage of self driving trucks will go hand in hand with the use of real-time data to optimize both efficiency and productivity of the service delivered, as a way to tackle traffic congestion for example. Driverless trucks could enable new business models that would see deliveries shift from day time to night time or time slots in which traffic is less heavily dense.

Suppliers

Company	Details
Waymo Semi	In March 2018, Waymo, the automated vehicle company spun off from Google parent company Alphabet Inc., announced it was applying its technology to semi trucks. In the announcement, Waymo noted it would be using automated trucks to move freight related to Google's data centers in the Atlanta, GA area. The trucks will be manned and operated on public roads.
Uber Semi	In October 2016, Uber completed the first driverless operation of an automated truck on public roads, delivering a trailer of Budweiser beer from Fort Collins, CO to Colorado Springs. The run was completed at night on Interstate 25 after extensive testing and system improvements in cooperation with the Colorado State Police. The truck had a human in the cab but not sitting in the driver's seat, while the Colorado State Police provided a rolling closure of the highway. At the time, Uber's automated truck was based primarily on technology developed by Otto, which Uber acquired in August 2016. In March 2018, Uber announced it was using its automated trucks to deliver freight in Arizona, while also leveraging the UberFreight app to find and dispatch loads.
Embark Semi	In February 2018, Embark Trucks announced it had completed the first cross-country trip of an automated semi, driving 2,400 miles from Los Angeles, CA to Jacksonville, FL on Interstate 10. This followed a November 2017 announcement that it had partnered with Electrolux and Ryder to test its automated truck by moving Frigidaire refrigerators from El Paso, TX to Palm Springs, CA.
Tesla Semi	In November 2017 Tesla, Inc., owned by Elon Musk, revealed a prototype of the Tesla Semi and announced that it would go into production. This long-haul, electric semi-truck can drive itself and move in "platoons" that automatically follow a lead vehicle. It was disclosed in August 2017 that it sought permission to test the vehicles in Nevada.
Starsky Robotics	In 2017, Starsky Robotics unveiled its technology that allows to make trucks autonomous. Unlike its bigger competitors in this industry that aims to tackle Level 4 and 5 Autonomy, Starsky Robotics is aiming at producing Level 3 Autonomy trucks, in which the human drivers should be prepared to respond to a "request to intervene" in case anything goes wrong.
Pronto AI	In December 2018, Anthony Levandowski unveiled his new autonomous driving company, Pronto, which is building L2 ADAS technology for the commercial trucking industry. The company is based in San Francisco.

Trains

Main article: Automatic train operation

The concept for autonomous vehicles has also been applied for commercial uses, like for autonomous trains. The world's first driverless urban transit system is the Port Island Line in Kobe, Japan, opened in 1981. First self-driving train in UK was launched in London Thameslink route.

An example of an automated train network is the Docklands Light Railway in London.

Also see List of automated train systems.

Trams

In 2018 the first autonomous trams in Potsdam were trialed.

Automated guided vehicle

Main article: Automated guided vehicle

An automated guided vehicle or automatic guided vehicle (AGV) is a mobile robot that follows markers or wires in the floor, or uses vision, magnets, or lasers for navigation. They are most often used in industrial applications to move materials around a manufacturing facility or warehouse. Application of the automatic guided vehicle has broadened during the late 20th century.

Aircraft

Main article: Unmanned aerial vehicle

Aircraft has received much attention for automation, especially for navigation. A system capable of autonomously navigating a vehicle (especially aircraft) is known as autopilot.

Delivery drones

Main article: Delivery drone

Various industries such as packages and food experimented with delivery drones. Traditional and new transportation companies are competing in the market. For example, UPS Flight Forward, Alphabet Wing, and Amazon Prime Air are all developing delivery drones. Zipline, an American medical drone delivery company, has the largest active drone delivery operations in the world, and its drones are capable of Level 4 autonomy.

However, even if technology seems to allow for those solutions to function correctly as various tests of various companies show, the main throwback to the market launch and use of such drones is inevitably the legislation in place and regulatory agencies have to decide on the framework they wish to take to draft regulation. This process is in different phases across the world as each country will tackle the topic independently. For example, Iceland's government and departments of transport, aviation, police have already started issuing licenses for drone operations. It has a permissive approach and together with Costa Rica, Italy, the UAE, Sweden and Norway, has a fairly unrestricted legislation on commercial drone use. Those countries are characterized by a body of regulation that may give operational guidelines or require licensing, registration and insurance.

On the other side, other countries have decided to ban, either directly (outright ban) or indirectly (effective ban), the use of commercial drones. The RAND Corporation thus makes the difference between countries forbidding drones and those that have a formal process for commercial drone licensing, but requirements are either impossible to meet or licenses do not appear to have been approved. In the US, UPS is the only one with the Part 135 Standard certification that is required to use drones to deliver to real customers.

However, most countries seem to be struggling on the integration of drones for commercial uses into their aviation regulatory frameworks. Thus, constraints are placed on the use of those drones such as that they must be operating within the visual line of sight (VLOS) of the pilot and thus limiting their potential range. This would be the case of the Netherlands and Belgium. Most countries do let pilot operate outside the VLOS but is subject to restrictions and pilot ratings, which would be the case of the US.

The general trend is that legislation is moving fast and laws are constantly being reevaluated. Countries are moving towards a more permissive approach but the industry still lacks infrastructures to ensure the success of such a transition. To provide safety and efficiency, specialized training courses, pilot exams (type of UAV and flying conditions) as well as liability management measures regarding insurances have to be developed.

There is a sense of urgency that breathes from this innovation as competition is high and companies lobby to integrate them rapidly in their products and services offerings. Since June 2017, the US Senate legislation reauthorized the Federal Aviation Administration and the Department of Transportation to create a carrier certificate allowing for package deliveries by drones.

Watercraft

Main article: Unmanned surface vehicle

Autonomous boats can provide security, do research, or perform hazardous or repetitive tasks (such as guiding a large ship into a harbor or transporting cargo).

Sea Machines

Sea Machines offers an autonomous system for workboats. While it does require a human operator to oversee its actions, the system takes care of a lot of active domain perception and navigation duties that normally a few members of the crew would have to do. They use AI to have situational awareness for different ships within the route. They utilize camera, lidar, and proprietary software to inform the operator of its status.

Buffalo Automation

Buffalo Automation, a team formed from the University of Buffalo, creates technology for semi-autonomous features for boats. They started out creating navigation assist technologies for freighters called AutoMate, which is like having another very experienced “first mate” that will look out for the ship.^[143] The system helps make twists and turns of difficult waterways.

Autonomous Marine Systems

This Massachusetts based company has led the forefront of unmanned sailing drones. The Datamarans are out autonomously sailing around to collect ocean data. They are created to enable large payload packages. Due to the automated system and their solar panels, they are able to navigate for longer periods of time. More than anything they boast their technologies on advanced metocean surveys which collect “wind velocity profiles with altitude, water current, conductivity, temperature profiles with depth, hi-resolution bathymetry, sub-bottom profiling, magnetometer measurements.” 

Mayflower

The autonomous vessel called Mayflower is expected to be the first large ship that makes an unmanned transatlantic journey.

Saildrones

This autonomous unmanned vessel uses both solar and wind energy to navigate.

DARPA

Sea Hunter is an autonomous unmanned surface vehicle (USV) launched in 2016 as part of the DARPA Anti-Submarine Warfare Continuous Trail Unmanned Vessel (ACTUV) program.

Submersibles

Main article: Autonomous underwater vehicle

Underwater vehicles have been a focus for automation for tasks such as pipeline inspection and underwater mapping.

Assistance robots

Spot

This robot is a four-legged nimble robot that was created to be able to navigate through many different terrain outdoors and indoors. It can walk on its own without colliding into anything. It utilizes many different sensors, including 360 vision cameras and gyroscopes. It is able to keep its balance even when pushed over. This vehicle, while it is not intended to be ridden, can carry heavy loads for construction workers or military personnel through rough terrain.

Highway Code change

The UK considers the way to update its british Highway Code for automated code:

Automated vehicles can perform all the tasks involved in driving, in at least some situations. They differ from vehicles fitted with assisted driving features (like cruise control and lane-keeping assistance), which carry out some tasks, but where the driver is still responsible for driving. If you are driving a vehicle with assisted driving features, you MUST stay in control of the vehicle.

— proposed changes to The Highway Code

If the vehicle is designed to require you to resume driving after being prompted to, while the vehicle is driving itself, you MUST remain in a position to be able to take control. For example, you should not move out of the driving seat. You should not be so distracted that you cannot take back control when prompted by the vehicle.

— proposed changes to The Highway Code

Concerns

Lack of control

Through the autonomy level, it is shown that the higher the level of autonomy, the fewer control humans have on their vehicles (highest level of autonomy needing zero human interventions). One of the few concerns regarding the development of vehicular automation is related to the end-users’ trust in the technology that controls automated vehicles. According to a nationally conducted survey made by Kelley Blue Book (KBB) in 2016, it is shown that the majority of people would still choose to have a certain level of control behind their own vehicle rather than having the vehicle operate in Level 5 autonomy, or in other words, completely autonomous. According to half of the respondents, the idea of safety in an autonomous vehicle diminishes as the level of autonomy increases. This distrust of autonomous driving systems proved to be unchanged throughout the years when a nationwide survey conducted by AAA Foundation for Traffic and Safety (AAAFTS) in 2019 showed the same outcome as the survey KBB did in 2016. AAAFTS survey showed that even though people have a certain level of trust in automated vehicles, most people also have doubts and distrust towards the technology used in autonomous vehicles, with most distrust in Level 5 autonomous vehicles. It is shown by AAAFTS’ survey that people's trust in autonomous driving systems increased when their level of understanding increased.

Malfunctions

A prototype of an autonomous Uber car being tested in San Francisco

        The possibility of autonomous vehicle's technology to experience malfunctions is also one of the causes of user's distrust in autonomous driving systems. In fact, it is the concern that most respondents voted for in the AAAFTS survey. Even though autonomous vehicles are made to improve traffic safety by minimizing crashes and their severity, they still caused fatalities. At least 113 autonomous vehicle related accidents have occurred until 2018. In 2015, Google declared that their automated vehicles experienced at least 272 failures, and drivers had to intervene around 13 times to prevent fatalities. Furthermore, other automated vehicles’ manufacturers also reported automated vehicles’ failures, including the Uber car incident. The self-driving Uber car accident that happened in 2018 is one of the examples of autonomous vehicle accidents that are also listed in List of self-driving car fatalities.  One of the reports made by the National Transportation Safety Board (NTSB) showed that the self-driving Uber car was unable to identify the victim in a sufficient amount of time for the vehicle to slow down and avoid crashing into the victim.

Ethical

Another concern related to vehicle automation is its ethical issues. In reality, autonomous vehicles can encounter inevitable traffic accidents. In situations like that, many risks and calculations need to be made in order to minimize the amount of damage the accident could cause. When a human driver encounters an inevitable accident, the driver will take a spontaneous action based on ethical and moral logic. However, when a driver has no control over the vehicle (Level 5 autonomy), the system of an autonomous vehicle is the one who needs to make that instant decision. Unlike humans, autonomous vehicles don't have reflexes and it can only make decisions based on what it is programmed to do. However, the situation and circumstances of accidents differ from one another, and one decision might not be the best decision for certain accidents. Based on two research studies in 2019, the implementation of fully automated vehicles in traffic where semi-automated and non-automated vehicles are still present might lead to many complications. Some flaws that still need consideration include the structure of liability, distribution of responsibilities, efficiency in decision making, and the performance of autonomous vehicles with its diverse surroundings. Still, researchers Steven Umbrello and Roman V. Yampolskiy propose that the value sensitive design approach is one method that can be used to design autonomous vehicles to avoid some of these ethical issues and design for human values.