Customer Testimony : Driver-Vehicle Cooperation for Automated Driving

By Polytechniques University - Hauts de France

CONTEXT – The development of automated driving, and more generally of advanced driver assistance systems interacting with the driver at the vehicle control level, generates problems of human-machine cooperation. Solving them requires to study, from the early design of the system, the ways the driver will interact with the system by addressing the problems of task sharing, authority, conflict management, level of automation and HMI. Works are done at the lab in order to implement these systems on demonstrators, in driving simulators and on real vehicles, and also to experimentally validate the developed systems.

 

CHALLENGE

To develop a cooperative architecture between drivers and assistance systems for automated driving, considering the driver’s state (in terms of vigilance, distraction, etc.)

SOLUTION

Implementation of a haptic shared control of the vehicle, coupled with a driver monitoring system in a multilevel architecture.

NEXT STEPS

The validation of the system in a simulator-based experiment has demonstrated its value, but it has also revealed some limitations. That is why the continuation of this work aims to integrate self-learning capacities in order 1) to allow the system to adapt its behaviour to the driver’s preferences, for acceptability and confidence matters, and 2) to extend its “skills” by observing the driver’s behaviour while coping with complex situations in manual driving mode.

BENEFITS

Such systems implementing a shared control have several advantages: They avoid frequent switching between automated and manual driving modes, they guarantee a good acceptability of the drivers, and they comply to the current regulations about “autonomous driving”.

 

The work realized at the lab allowed to develop a cooperative architecture between drivers and assistance systems for automated driving (sharing of vehicle control), to implement these systems on demonstrators in driving simulators and on real vehicles, and to experimentally validate the systems developed.

More precisely, this work comprised these different steps:

  • To define a multi-level cooperation architecture allowing the interactions between the driver and the autonomous vehicle both at the vehicle control level (vehicle guidance on the defined trajectory) and at the tactical level (choice of different possible manoeuvres: change of lane, insertion, overtaking, etc.);
  • To identify the information the driver needs to carry out his driving task and this according to the level of automation (shared driving, supervision, delegation, takeover);
  • To design the HMI mechanisms to provide this information to the driver and to collect his instructions (visual, manual, sound, haptic GUIs), depending on the context and mode involved;
  • To define the Driver Monitoring (DM) needs during both the manual and the delegated driving phases in order to ensure the safety of the vehicle;
  • To define the mechanisms of transitions between driving modes and the conditions associated with these transitions in order to ensure safety;
  • To prototype all this work in the form of an integrated system on a dynamic interactive driving simulator;
  • To prototype part of this work on a real vehicle;
  • To evaluate the prototype on the driving simulator in the context of an experiment involving a representative panel of drivers.

This project demonstrated the need for close cooperation between the driver and the assistance system in order to guarantee the acceptability and safety of the autonomous vehicle. For this, a multi-level cooperation architecture was developed and prototyped on a driving simulator. The algorithms implemented in this architecture to solve the problems of control sharing, authority management, conflict resolution for decision-making have highlighted the need for a better knowledge of the “operation” of the driver.

Several axes emerged from this work:

  • The need for models of the driver describing both its decision-making and sensory-motor mechanisms. The models of the decision-making mechanisms that lead the driver to choose an alternative rather than another will make the behaviour of the automatism with that of the driver homogeneous in order to favour the acceptability of the systems. They will also advance the problem of ethics in the field of the autonomous vehicle. The sensorimotor models will allow to refine the interfaces, especially haptic, between driver and system, for a better safety.
  • The need for advanced driver monitoring. In many driving phases, it is necessary to relate the driver’s operational capabilities to the requirements of the situation. Vigilance, workload, situation awareness, attention, etc. are all essential information for the modulation of the level of automation.

“The implementation of a shared control of the vehicle under the supervision of a driver monitoring system allow to develop systems compliant with the current regulation and with a good acceptability for the driver. Self-learning capacities allow the system to match the preferences of the driver and also to extend its “skills” in terms of manoeuvres, by observing the driver during the manual and shared driving phases”.