ASSISTANCE SYSTEM FOR EVASIVE MANOEUVRES ON CURVES OPTIMAL COPILOT FOR VEHICLE CONTROL AT THE LIMITS OF FRICTION
- Delivery
- Available on this site
- Format
- Price
- Non-members (tax incl.):¥1,100 Members (tax incl.):¥880
- Publication code
- 20219064
- Paper/Info type
- Other International Conferences
- Pages
- 1-6(Total 6 p)
- Date of publication
- Sep 2021
- Publisher
- JSAE
- Language
- English
Detailed Information
Author(E) | 1) Mathias Lidberg, 2) Dalila Avdic, 3) Timothy Gordon |
---|---|
Affiliation(E) | 1) Chalmers University of Technology, 2) University of Lincoln, 3) University of Lincoln |
Abstract(E) | This paper uses optimal control methods to determine a control algorithm for assisting drivers with evasive manoeuvres. On a curved road, an evasive lane change requires a critical balance between avoiding the obstacle, recovering the curved path, and maintaining a suitable degree of yaw stability throughout the manoeuvre. Fully automated collision avoidance by steering is highly problematic, due to the high-risk collision with an undetected vehicle or other objects. So, the aim here is to use object detection and vehicle state estimation as inputs to an integrated chassis controller, but leave the decision making to the driver. When an obstacle is detected by both the driver and the ADAS system, the driver’s steering action triggers a system response – individual wheel braking is applied to assist the driver. To determine an ideal response, offline numerical optimization is performed in the software PROPT, this being applied to a combined driver-vehicle model. This benchmark is then compared to a real-time feasible sub-optimal controller based on the Modified Hamiltonian Algorithm (MHA). This method uses a friction-limited particle motion reference. To demonstrate the capabilities of the method, the scenario tested is for a low-friction surface, with the speed and curvature such that the vehicle is cornering close to the limiting lateral acceleration before the obstacle is detected. The MHA controller is found to give a good approximation to the PROPT benchmark. The method shown is very general and can be applied in a wide variety of ADAS scenarios. |