Autonomous Vehicles: Moving on up

Roberto Ponticelli from Horiba Mira explains the challenges with increasing levels of autonomy and highlights how engineers are working to overcome these issues
At first glance, the levels of vehicle autonomy appear simple – a clearly defined set of increasingly autonomous systems. The widely used SAE levels 0 to 5 provide a gradient from no autonomy where the human driver monitors the driving environment, up to full autonomy where an automated driving system monitors the driving environment.
It appears simple enough, but when looking deeper, these levels become much more complex. Looking beyond the surface, this classification system allows for a vehicle to be classified at one level under certain driving conditions or certain features, and classified at a different level for other conditions and features. For example, a vehicle with anti-lock braking would be classified as level 1, while the same vehicle with adaptive cruise control would be level 2.
This complexity means vehicle manufactures are not required to take a linear approach in developing autonomous vehicles. In fact, some manufacturers are skipping levels entirely, such as Ford and Google. Ford hit the headlines earlier this year when they made the decision to skip level 3, citing the handover of control to the driver as the main challenge. Before this, Google opted to target the highest level of autonomy for its Waymo project for the same reason. This begs the question, why have a level 3 if it is deemed too complicated?
There are several challenges posed by each level, including driver handover, consumer acceptance, electromagnetic compatibility, cyber security and skills. These challenges are being tackled head on by engineers across the globe who are working towards the safe development of vehicles at each level of autonomy.
Driver handover
At level 3 and above, the human driver is less a driver and more a passenger. When the driver is removed from the task of actively monitoring the road and responding to changes in the external environment, the ability for them to take immediate control of the car is difficult. The brain takes time to recognise a change, such as a new fault in the vehicle that prevents it from carrying out its designed task. Being able to recognise this fault and actively switch from passenger to driver takes valuable time.
Engineers are working hard to overcome this, such as devising simulators for vehicle control systems where the vehicle can be driven safely without exposing the public to the dangers of a work in progress. Private facilities to develop and test prototype vehicles also remain intrinsic to the success of developing autonomous systems safely.
Electromagnetic compatibility
In addition to driver handover, the increasing number of software systems in each autonomous level has a significant impact on electromagnetic compatibility. Wherever there is an electronic unit, there exists the possibility of interfering with its operation.
Depending on the system’s integrity, its susceptibility to fail under certain conditions can vary. Everyone is familiar with the effects of a mobile phone when placed near a live microphone, and while the electromagnetic interference is easy enough to fix, with complex autonomous systems, the challenge multiplies.
The demand for connected and autonomous cars is also causing changes in the electromagnetic environment in which vehicle control systems must operate. This includes commonly used technologies, such as wifi and the vehicular short-range communications (DSRC, ITS-G5, Wave) in the 5GHz band, which are now common in most premier vehicles.
This is driving changes to both the susceptibility test frequencies and emissions bands included in EMC standards. This has the knock-on effect of increasing the bandwidth and upper frequency of the EMC test antennas and the measurement and capability requirements of test receivers and power amplifiers.
Cyber security
In addition to electromagnetic compatibility issues, cyber security and the threat of hacking poses another challenge for each increasing level of autonomy.
A motivated attacker can identify and exploit a vulnerability via a number of entry points. Examples include wireless interfaces such as cellular, Bluetooth and keyless entry systems, wired connections such as the OBD-II diagnostic port, sensors and attacks on the electronic control units themselves. New methods to attack embedded systems are continuously being developed, leading to a cat-and-mouse battle between those developing the attacks and those developing the defences. This means that as well as taking account of known attacks and defending against known vulnerabilities, a focused R&D effort is necessary to search for unknown attacks and potential weaknesses and to develop appropriate security measures to protect against them.
Consumer acceptance
Consumer perception and acceptance of autonomous vehicles is still in a delicate position with only half of the population in the UK, USA and Australia positive about autonomous technology according to a University of Michigan Transportation Research Institute survey.
To counter this perception, the automotive industry will need to work hard to educate the public on the benefits and safety of autonomy. Aeroplanes are now flown predominately using auto-pilot, which is widely accepted by passengers as a safe mode of transport. This acceptance didn’t happen overnight, and neither will the acceptance of autonomous vehicles.
Skills
The automated vehicle world is similar to that of robotics, with perhaps a stronger consciousness of the need for high levels of safety and reliability at every level of the value chain, from concept to design and manufacturing to verification. This means the development of autonomous vehicle systems requires niche skills that aren’t traditionally seen within software engineering departments. These skills include computing science knowledge, from artificial intelligence and safety-critical software development to software cyber security and system simulations, as well as knowledge of human-factors related to mobility.
Human body
Human factors in relation to mobility are essential as they call into question the overall safety of a vehicle and its internal systems. The knowledge of what happens to the human body as a vehicle travels through air is widely studied – when you accelerate quickly or brake quickly, your body experiences a range of symptoms such as changes in blood pressure, shifting of organs and physical bodily movement.
With autonomous vehicles, these challenges increase with the rise of software systems that emit electromagnetic fields such as RF. RF can cause tissue heating, which in high doses can be harmful.
A number of national and international bodies have therefore recommended limits for human exposure to electromagnetic fields (for example, 1999/519/EC), which also include methods for assessing the risks associated with exposure to multiple sources. Although these limits are often interpreted as requirements on the characteristics of individual equipment, they in fact apply to the overall electromagnetic environment to which humans may be exposed.
Consequently, the exposure levels that can be tolerated from any specific piece of equipment are not absolute values, but depend on the levels due to other sources that can be expected to be present in its intended operating environment.
Conclusion
With each level of autonomy, each of these issues inevitably increases. The complexity of vehicle systems combined with the human factor in terms of acceptance and the ability to respond in the failure of a vehicle system shed more light into the reasons for selecting certain autonomous levels over others. That said, there is a clear place for each level and while some manufacturers have chosen to aim for the highest, there will always be a place for the lower levels of autonomy.
Roberto Ponticelli is chief engineer for intelligent mobility at Horiba Mira
