It is 2023, we do not yet have flying cars nor do we have fully autonomous cars plying the roads. From established automakers like Mercedes Benz to upstarts like Tesla, automobile companies seem to be struggling to introduce a truly autonomous driving system that is safe. But now, an international team of researchers are proposing an alternative to fully autonomous driving technologies—a definitive collaboration between humans and machines.
According to École polytechnique fédérale de Lausanne (EPFL), available data on automated driving shows that placing too much control of a vehicle in the hands of automation can do more harm than good. This is because human drivers get careless and overly rely on autonomous driving systems when they are fully automated.
In order to develop a system that is a good compromise between a manual vehicle and a fully autonomous vehicle, researchers from EPFL partnered with Japanese steering system supplier JTEKT to develop and road-test a haptics-based automated driving system that “collaborates with the driver.” The research was published last month in the journal Nature Communications Engineering.
“The mass deployment of fully automated driving vehicles, reliable in any situation, is very challenging and likely not to happen in the coming five years. The driver cannot simply be taken out of the equation. From the perspective of the experience gained in the market with Level 2 vehicles, the ‘out of the loop problem’ or misuse of driver while operating ADAS (advanced driver assistance systems) functions reveals the still low maturity of the existing technology. Understanding ’driver engagement’ is of critical importance for safety and social acceptance,” JTEKT researcher Tomohiro Nakade told indianexpress.com over email.
Nakade compares the new driving system developed by EPFL and JTEKT to how a horseback rider communicates with a horse using reins. Instead of a driver completely controlling a vehicle or the vehicle controlling itself, they both work together.
Collaborative steering system: How it works
The collaborative steering system developed by the researchers integrates information from the vehicle’s steering column apart from data from cameras, radar, and LIDAR. The system continuously encourages engagement between the driver and the system, unlike most current autonomous driving systems, which are either on or off and nothing in between.
According to Robert Fuchs, who is part of the research team, humans lose the ability to react when they are just monitoring an autonomous system, instead of being actively involved in the process. This is why the researchers wanted to increase driver engagement through automation.
The system’s working can broadly be classified into three functionalities—interaction, arbitration and inclusion. The system begins by distinguishing between four different types of human-robot interactions—cooperation, coactivity, collaboration and competition.
Identifying the interaction
Cooperation refers to when the human and the automation system are working together towards one goal, like how current electric power steering systems help reduce the steering load by boosting the driver’s input during manual operation. Co-activity refers to when the human and the system have different goals but their actions impact each other, like with current lane assistance systems that keep drivers centred within the lane they are driving in.
Collaboration is when the human and automation system assist each other in achieving different goals. It is a more driver-oriented approach where for example, if the system concedes steering authority to the driver, it remains ready to back up the driver even when it is inactive.
Competition refers to the extreme cases when the driver and the system act in opposition to each other. “Although it is an extreme case, it can be implemented in the form of “automated emergency steering” when the driver cannot prevent an imminent collision to ensure safety. Competition should not be seen strictly as fully rejecting the driver. It can also be used to gently oppose the driver when proceeding to a dangerous manoeuvre, e.g. steering towards a guard rail,” explained Nakade.
Arbitrating between interaction modes
After identifying the appropriate interaction mode for the moment, the system “arbitrates” or switches between the various modes. For example, the system could be in co-activity mode as it helps to keep the driver at the centre of the lane as the vehicle is being driven on the highway when suddenly, it detects an impending collision.
At this point, the driver may not react fast enough to avoid the collision, meaning that the system will have to oppose what the driver is doing. For this, the system will switch from co-activity mode to competition mode, to oppose the driver’s actions and ensure that the vehicle does not crash.
Inclusion of the driver and their inputs
The final and perhaps crucial part of the system is how it includes the driver and their inputs and every juncture. Unlike other autonomous driving systems, if the user intervenes by turning the steering wheel, the collaborative driving system does not consider this to be an “override.” Instead, it will recalculate the vehicle’s trajectory and include it in its measurements.
Testing the new collaborative steering system
After developing the system, the next step was to test and validate it to understand how it affects the driver’s safety and comfort. To test it, the researchers conducted many experiments that involved a simulated virtual driver and a human driver using a detached power steering system. Apart from tests carried out on a driving simulator, the researchers also carried out field tests with a modified test vehicle.
The field tests were conducted with the help of five drivers who drove on a JTEKT test course in Japan, where the researchers connected the new system to a standard sedan. During the tests, the researchers paid special attention to testing the driver’s experiences of steering smoothness and ease of lane-changing. The tests confirmed that the system holds significant potential to increase comfort while reducing effort for drivers.
Based on the academic foundations of the technology published in the research paper, JTEKT in October last year launched “Pairdriver”, a human-centered automated steering control system. For future revisions to the technology, the researchers envision adding more channels of feedback apart from haptics while also improving the individual technologies that are part of the system.
(Image: Tesla) 
Share
