Human-like overtaking maneuvers using Inverse Optimal Control

Human-like overtaking maneuvers using Inverse Optimal Control

van Weperen, Marijke (TU Delft Mechanical, Maritime and Materials Engineering; TU Delft Cognitive Robotics)

Abbink, David (mentor)
Alonso Mora, Javier (mentor)

Delft University of Technology

Mechanical Engineering | Systems and Control

2019-06-19

In recent years, rapid progress has been made towards automated highway driving. To increase comfort and trust in these automated cars, their motion planners, such as Model Predictive Controllers (MPCs), should exhibit a human-like driving style. However the effect of surrounding traffic on overtaking maneuvers has not yet been investigated. Here, the influence of the relative speed between cars and the width of the to-be-overtaken car on overtaking trajectories is analysed. In a driving simulator experiment cars with different widths (normal: 2m, wide: 2.8m) were overtaken at different relative speeds (20km/h, 50km/h and free speed between 20-50km/h). Drivers were found to decrease their safety time margin, measured as minimal Time To Collision (TTC), for increasing relative speeds from 20km/h (normal: Mean (M) = 6.74s, wide: M = 7.18s) to 50km/h (normal: M = 3.89s, wide: mean = 4.69s) and for decreasing car width. Additionally, increasing car width of the other car decreases the road margin at a relative speed of 50km/h (normal: M = 1.01m, wide: M = 0.91m). Due to large between-subject variability, individualized adaptation strategies of the controller are needed. An offline optimization method called Inverse Optimal Control (IOC) optimizes a personalized cost function for the MPC, based on demonstrated inputs and outputs, in about 1 hour. The lateral reference of the MPC adapts to different relative speeds and widths of the other car. By tracking the reference, the motion planner generates human-like trajectories online similar to the demonstrated ones within a small error (M = 0.17m)

Inverse Optimal Control
Overtaking
Human-like driving

http://resolver.tudelft.nl/uuid:ac18235e-6ef1-4ca0-bf37-dd70b12bc252

Student theses

master thesis

© 2019 Marijke van Weperen