Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Safe Hybrid-Action Reinforcement Learning-Based Decision and Control for Discretionary Lane Change

Ruichen Xu, Xiao Liu, Jinming Xu, Yuan Lin

TL;DR

This work introduces safe hybrid-action reinforcement learning into discretionary lane change for the first time and proposes the Parameterized Soft Actor–Critic with PID Lagrangian (PASAC-PIDLag) algorithm, which is an unsafe version of PASAC-PIDLag.

Abstract

Autonomous lane-change, a key feature of advanced driver-assistance systems, can enhance traffic efficiency and reduce the incidence of accidents. However, safe driving of autonomous vehicles remains challenging in complex environments. How to perform safe and appropriate lane change is a popular topic of research in the field of autonomous driving. Currently, few papers consider the safety of reinforcement learning in autonomous lane-change scenarios. We introduce safe hybrid-action reinforcement learning into discretionary lane change for the first time and propose Parameterized Soft Actor-Critic with PID Lagrangian (PASAC-PIDLag) algorithm. Furthermore, we conduct a comparative analysis of the Parameterized Soft Actor-Critic (PASAC), which is an unsafe version of PASAC-PIDLag. Both algorithms are employed to train the lane-change strategy of autonomous vehicles to output discrete lane-change decision and longitudinal vehicle acceleration. Our simulation results indicate that at a traffic density of 15 vehicles per kilometer (15 veh/km), the PASAC-PIDLag algorithm exhibits superior safety with a collision rate of 0%, outperforming the PASAC algorithm, which has a collision rate of 1%. The outcomes of the generalization assessments reveal that at low traffic density levels, both the PASAC-PIDLag and PASAC algorithms are proficient in attaining a 0% collision rate. Under conditions of high traffic flow density, the PASAC-PIDLag algorithm surpasses PASAC in terms of both safety and optimality.

Safe Hybrid-Action Reinforcement Learning-Based Decision and Control for Discretionary Lane Change

TL;DR

This work introduces safe hybrid-action reinforcement learning into discretionary lane change for the first time and proposes the Parameterized Soft Actor–Critic with PID Lagrangian (PASAC-PIDLag) algorithm, which is an unsafe version of PASAC-PIDLag.

Abstract

Autonomous lane-change, a key feature of advanced driver-assistance systems, can enhance traffic efficiency and reduce the incidence of accidents. However, safe driving of autonomous vehicles remains challenging in complex environments. How to perform safe and appropriate lane change is a popular topic of research in the field of autonomous driving. Currently, few papers consider the safety of reinforcement learning in autonomous lane-change scenarios. We introduce safe hybrid-action reinforcement learning into discretionary lane change for the first time and propose Parameterized Soft Actor-Critic with PID Lagrangian (PASAC-PIDLag) algorithm. Furthermore, we conduct a comparative analysis of the Parameterized Soft Actor-Critic (PASAC), which is an unsafe version of PASAC-PIDLag. Both algorithms are employed to train the lane-change strategy of autonomous vehicles to output discrete lane-change decision and longitudinal vehicle acceleration. Our simulation results indicate that at a traffic density of 15 vehicles per kilometer (15 veh/km), the PASAC-PIDLag algorithm exhibits superior safety with a collision rate of 0%, outperforming the PASAC algorithm, which has a collision rate of 1%. The outcomes of the generalization assessments reveal that at low traffic density levels, both the PASAC-PIDLag and PASAC algorithms are proficient in attaining a 0% collision rate. Under conditions of high traffic flow density, the PASAC-PIDLag algorithm surpasses PASAC in terms of both safety and optimality.
Paper Structure (16 sections, 19 equations, 8 figures, 4 tables, 1 algorithm)

This paper contains 16 sections, 19 equations, 8 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Reinforcement Learning Preliminaries
  3. Soft Actor-Critic
  4. Parameterized Soft Actor-Critic
  5. Parameterized Soft Actor-Critic with PID Lagrangian
  6. Lane Change Problem Formulation
  7. Lane Change Environment
  8. Environment State
  9. Control Action
  10. Reward and Cost
  11. Experiments and Results
  12. Training
  13. Testing
  14. Comparison and Analysis
  15. Generalization Analysis
  16. ...and 1 more sections

Figures (8)

  • Figure S1: Lane change environment created using SUMO, with the ego vehicle depicted in red and other traffic participants in green.
  • Figure S2: The training progress of the PASAC-PIDLag algorithm compared to the PASAC algorithm.
  • Figure S3: The vehicle's velocity, acceleration, and distance of the leader vehicle under the regulation of PASAC-PIDLag algorithmic controls. A black dashed line traverses the graphs, symbolizing the execution of a successful lane change by the ego vehicle.
  • Figure S4: The figure illustrates a successful lane change maneuver executed by a vehicle under the control of the PASAC-PIDLag algorithm, where the red vehicle is denoted as the ego car, and the green vehicles represent the surrounding traffic.
  • Figure S5: The velocity, acceleration, and lead vehicle distance during a collision event due to lane changing under the PASAC algorithm.
  • ...and 3 more figures