A Collision Relationship-Based Driving Behavior Decision-Making Method for an Intelligent Land Vehicle at a Disorderly Intersection via DRQN

Abstract

An intelligent land vehicle utilizes onboard sensors to acquire observed states at a disorderly intersection. However, partial observation of the environment occurs due to sensor noise. This causes decision failure easily. A collision relationship-based driving behavior decision-making method via deep recurrent Q network (CR-DRQN) is proposed for intelligent land vehicles. First, the collision relationship between the intelligent land vehicle and surrounding vehicles is designed as the input. The collision relationship is extracted from the observed states with the sensor noise. This avoids a CR-DRQN dimension explosion and speeds up the network training. Then, DRQN is utilized to attenuate the impact of the input noise and achieve driving behavior decision-making. Finally, some comparative experiments are conducted to verify the effectiveness of the proposed method. CR-DRQN maintains a high decision success rate at a disorderly intersection with partially observable states. In addition, the proposed method is outstanding in the aspects of safety, the ability of collision risk prediction, and comfort.

Sensor Fusion and Advanced Controller for Connected and Automated Vehicles

“…In addition, an efficient and high-precision estimation framework for a Four-Wheel Independently Actuated (FWIA) autonomous vehicle was focused on [ 3 ], and an efficient measurement method for brake pressure change rate was reported [ 4 , 5 ]. In the decision-making and path-planning layers, a decision-making framework based on the Extended Collision Warning System (ECWS) [ 6 ] and a collision relationship-based decision-making method based on the Deep Recurrent Q Network (CR-DRQN) [ 7 ] were proposed. Furthermore, a path planning method based on the Takagi–Sugeno (TS) fuzzy-model-based approach [ 8 ] and a motion planning method by minimizing motion sickness [ 9 ] were proposed.…”

“…In the study [ 7 ], a collision relationship-based driving behavior decision making method for intelligent land vehicles was presented, which successfully solved the problem of instability in decision-making caused by decreased perceptual confidence. This method is based on the approach of the Deep Recurrent Q Network (DRQN), which is a combination of the Deep Q Network (DQN) and the Long-Short Term Memory (LSTM).…”