Your regular update for technical and industry information
Your regular update for technical and industry information
The deployment of electronic control systems, such as anti-lock brakes or electronic stability programs, has contributed to a huge reduction in the consequences of traffic accidents. Further potential is associated with the gradual automation of driving, that is, the progressive handover of driving tasks to electronic systems. The culmination of this process is robotised vehicles, which will not require a driver under any circumstances. This trend already began some years ago and is moving forward with each new generation of vehicles.
Advanced driver assistance systems - whether high beams assistant or autonomous emergency braking - are taking over more and more activities from the driver, helping in common situations, warning of danger, preparing the driver for accidents or supporting the driver in critical situations.
The development of assistance systems is accompanied by the development in the monitoring of the vehicle itself and its surroundings. While electronic stability systems should monitor internal vehicle states with on-board sensors, the collision avoidance system or traffic sign recognition requires information about the vehicle’s surroundings based on these on-board sensors like cameras. The next step will be to link vehicles with infrastructure and other vehicles, allowing for a much wider range of information than the range of the vehicle on-board sensors.
If the advanced driver assistance systems that take over driving tasks are to be successful, the equipment in vehicles must meet a fairly ambitious goal: to be safe enough. However, it is still not clear what this means for the development and testing of such systems.
Increasing of safety, however, is not the only objective of driver assistance systems. In terms of the day-to-day operation, growing comfort of vehicle control is also important. And this convenience will gradually increase, until automatic systems completely take over driving and no driver will be needed. But that is the music of the distant future. We will first encounter automated driving in a simple enclosed environment, for example, a logistics centre or a car park. This will probably be followed by driving on selected motorways, on which the infrastructure will be guaranteed, and then urban roads will have their turn. In the case of commercial vehicles, the term platooning can be heard increasingly often, referring to the driving of trucks in a mutually communicating platoon on a motorway, while only the first vehicle is driven by a driver, the others are controlled electronically.
The speed of penetration of advanced driver assistance systems in the lower category vehicles is astonishing. The first systems to measure the distance between vehicles and, in today’s terms “merely” warn the driver, appeared in 1992. The first vehicle with adaptive cruise control, i.e. a system that controls both acceleration and braking, appeared in 1999; this system was then extended to include an autonomous emergency braking (AEB) function in 2003.
The table represents the levels of assistance systems. Nowadays, we are increasingly seeing level 1 and 2 systems in usual vehicles. The level zero vehicles are disappearing from the European market, due to pressure from consumer organisations such as Euro NCAP. With the lower level systems the full responsibility lies with the driver, and therefore it is more dangerous when the system reacts incorrectly when no reaction is expected. This situation is referred to as false-positive. A situation where the system is supposed to react, but does not react (false-negative) is not so critical in terms of its contribution to safety and acceptance. The human driver must supervise the system and is fully responsible in case of an accident.
The opposite situation occurs when driving a vehicle is taken over by an electronic control system in a certain situation or completely. Then the question arises as to whether the driver is still responsible for any accident. The situation described as false-negative can therefore be fatal to the occupants and their surroundings. A sun-blind camera or a snow-covered radar sensor will not be allowed to cause an immediate switching-off of the electronic control system. Another very important condition is the introduction of recording devices, so-called black boxes, which collect information on who was driving the vehicle at the given moment, i.e. whether the vehicle was in automatic mode or driven by the human driver before an accident.
There are no clear assessment standards for complex advanced driver assistance systems that cooperate with other parts of a vehicle, such as the engine, brakes or steering, or even communicate with other road users and infrastructure. How to evaluate their contribution? In terms of safety? According to the probability and severity of accidents? How to predict them? Is safety the only criterion? How to define the references? And what about the acceptance by the user? Research and development teams around the world are concerned with these and similar issues.
However, we can already imagine that the operation of autonomous vehicles will only be acceptable if they are safer than current ones, driven by human drivers. As is already noticeable, their deployment brings with it new types of accidents, but they should be less serious. The main benefit will be the reduction of fatal accidents. We can also talk about improving traffic flow and the associated reduction of emissions today. Last but not least, car driver comfort and truck driver productivity will increase. Standards are currently lacking not only for the development of vehicles and their components, but also for their approval.
NCAP (New Car Assessment Programme) consumer organisations evaluate advanced car assistance systems. Currently there are about nine different NCAP organisations in the world. Not all test AEB systems as of yet. In terms of testing procedures, the individual testing approaches of NCAP organisations are not uniform. While the American US NCAP prescribes tests for human drivers and their inaccuracies are resolved by a higher number of tests, the European Euro NCAP relies on steering and pedal robots for repeatability. However, it will include also subjective human machine interface assessemnt in the future. Precise position measurement is an essential prerequisite. The benefits of the European approach are clear: greater accuracy in terms of the position in the traffic lane and speed reduces the number of tests required and provides reproducible results. AEB systems are currently being tested in three areas in Euro NCAP consumer trials: AEB City, AEB Interurban and AEB Pedestrian. In 2018, cyclist are being added.
However, the Euro NCAP tests represent only a very limited part of the AEB systems workspace, in other words, the tests cannot guarantee the functionality of AEB in other situations. What may be enough for consumer tests is not sufficient as an effectiveness assessment. In addition, only situations in which the system is requested to react are being tested. The ‘false-positive’ situation, i.e. when the system reacts but it should not, is not on the Agenda for Euro NCAP.
The new series of UNECE Regulation No. 79, which already contains rules for the approval and testing of assisted driving, came into force for vehicles in 2017. New trucks and buses in the European Union have had to be equipped with new AEB and LDW systems since 2013. Their approval is carried out in accordance with UNECE Regulations No. 130 and 131.
For example, the timings and types of warnings together with the automatic braking process for the AEB approval are described in UNECE Regulation No. 131. Tests are carried out both with a stationary and a moving targets corresponding to a passenger car (category M1, class AA). The initial speed of trucks is 80 km/h. The AEB system first warns the driver and then it should brake automatically. Warning times vary between systems according to the categories N2 and N3; they are shorter for N2. Then the self-diagnostic is checked. Even the passage between two M1 category parked vehicles with a lateral distance of 4.5 m is tested. In this case, the system must not respond, as this is an example of the false-positive category.
In addition to consumer organisations, developers and testing organisations also define their own testing methodologies. Such tests are much more complex, but even so they will not be able to cover the whole range of possible situations. These methodologies evaluate ADAS systems from several perspectives, including human machine Interface or ride comfort.
Due to the mentioned complexity of systems from level 1 up and their workspace, a large number of completely different scenarios are required for verification and validation, trying to cover their Workspace as much as possible. Driving tests on a polygon will not be enough, since their length will significantly slow down the entire development of the vehicle. Simulation will therefore play a decisive role in the validation of advanced driver assistance systems and automated driving in the foreseeable future.
An example of AEB is a situation where a pedestrian enters a traffic lane between parked vehicles. Since the AEB works only within the bounds of physical laws, it is not possible to avoid an accident in all situations. The result is influenced by situational variants and parameters, such as the probability of vehicle sensors sensing pedestrians (sensor efficiency), vehicle speed, pedestrian speed, the likelihood that a pedestrian notices vehicles and does not step into the road, driver response time, adhesion or distance between the vehicle and pedestrian when entering the road and many others.
The basis is a combination of experimental methods and simulation. The simulation is based on the fact that various cases of critical situations can be described statistically. First, it is necessary to identify the cause of an accident where automated systems could be potentially beneficial. Let’s keep in mind, for example, that lack of attention associated with use of communication devices or driver's fatigue is a very common cause. In such situations, automated systems could have a significant advantage. In addition, the overall traffic situation must be taken into account. From the results of such a statistical analysis, we get the final number of test cases that can be examined by testing on dynamic simulators or directly on a proving ground combined with simulation. The aim of the simulation level is to create sets of scenarios of potentially critical situations. In the simulation environment, experts can selectively modify the parameters of the situation to test as many different scenarios as possible.
Not only the maturity of simulation technologies allows their application to demonstrate the proper functionality of advanced driver assistance systems and autonomous driving. In order to get clearer outlines for their deployment, various panels are at work and projects underway to develop approaches that allow a combination of simulation and proving ground experiments as a basis for verification and validation of the effectiveness of driver assistance and automated systems.