A Safe and Efficient Lane Change Decision-Making Strategy of Autonomous Driving Based on Deep Reinforcement Learning

Abstract

As an indispensable branch of machine learning (ML), reinforcement learning (RL) plays a prominent role in the decision-making process of autonomous driving (AD), which enables autonomous vehicles (AVs) to learn an optimal driving strategy through continuous interaction with the environment. This paper proposes a deep reinforcement learning (DRL)-based motion planning strategy for AD tasks in the highway scenarios where an AV merges into two-lane road traffic flow and realizes the lane changing (LC) maneuvers. We integrate the DRL model into the AD system relying on the end-to-end learning method. An improved DRL algorithm based on deep deterministic policy gradient (DDPG) is developed with well-defined reward functions. In particular, safety rules (SR), safety prediction (SP) module and trauma memory (TM) as well as the dynamic potential-based reward shaping (DPBRS) function are adopted to further enhance safety and accelerate learning of the LC behavior. For validation, the proposed DSSTD algorithm is trained and tested on the dual-computer co-simulation platform. The comparative experimental results show that our proposal outperforms other benchmark algorithms in both driving safety and efficiency.

A Comprehensive Review on Limitations of Autonomous Driving and Its Impact on Accidents and Collisions

“…These sensors enable vehicles to detect and respond to dynamic surroundings more accurately, minimizing the risk of collisions. Additionally, the use of high-definition mapping and localization systems assists in precise navigation, contributing to safer decision-making by the autonomous vehicle [156]. To improve safety, researchers Fig.…”

An overview of the development of key technologies in autonomous driving technology

“…and Pan M.Z. address the problem that existing driverless strategies ignore the driving logic that humans follow when driving a car, the authors impose the effect of the rule constraints of human-like driving on the successive behaviors of the intelligence in a deep reinforcement learning-based end-to-end driverless control network and build a driverless end-to-end control network capable of outputting successive ordered behaviors that conform to the logic of human-like driving [3]. The authors also reduced the output rate of dangerous behaviors of the control strategy by applying a posteriori feedback to the output.…”

How Do Autonomous Vehicles Decide?

“…RL plays an imperative role in the decision-making process of AD, which enables AVs to acquire an ideal driving strategy through constant interaction with the environment. Lv et al [ 84 ] developed a deep RL (DRL)-based motion planning technique for AD scenarios including an AV merging into two-lane road traffic flow and executing lane-changing maneuvers on highways. An improved version of the DRL algorithm based on a deep deterministic policy gradient (DDPG) is created with well-defined reward functions.…”