Deep Reinforcement Learning in Autonomous Car Path Planning and Control: A Survey

TL;DR

The paper surveys how Deep Reinforcement Learning ($DRL$) is applied to local path planning and control in autonomous driving, framing the problem within the $MDP$ formalism and surveying algorithms from value-based (e.g., $DQN$, Double DQN, Dueling DQN) and policy-based families (e.g., $PPO$, $TRPO$), as well as actor–critic methods (e.g., $DDPG$, $TD3$, $SAC$). It analyzes applications across trajectory planning, vehicle control, and end-to-end driving, highlighting how SAC, PPO, TD3, DQN, and DDPG are commonly used in planning tasks, while model-based DRL and adaptive dynamic programming (ADP) are emphasized for long-horizon control and integration with dynamics models. The survey discusses design choices for states, actions, and reward functions, the challenges of sample efficiency and training stability, and the simulation‑to‑real transfer via domain randomization and interpretable end‑to‑end approaches, such as the IDC and VISTA/SESR frameworks. It concludes with strategies to improve comparability, safety, and real‑world deployment, and outlines future directions including better benchmarks, safer training paradigms, more realistic simulators, and real‑time, compute‑efficient architectures to enable robust DRL‑driven autonomous driving. Overall, the work clarifies DRL’s potential to enhance path planning and control while identifying practical hurdles to scalable, safe, and transfer‑ready autonomous systems.

Abstract

Combining data-driven applications with control systems plays a key role in recent Autonomous Car research. This thesis offers a structured review of the latest literature on Deep Reinforcement Learning (DRL) within the realm of autonomous vehicle Path Planning and Control. It collects a series of DRL methodologies and algorithms and their applications in the field, focusing notably on their roles in trajectory planning and dynamic control. In this review, we delve into the application outcomes of DRL technologies in this domain. By summarizing these literatures, we highlight potential challenges, aiming to offer insights that might aid researchers engaged in related fields.

Figures (17)

• Figure 1: Control architecture for automated vehiclesstano2023model.
• Figure 2: Methodology of screening and selecting papers in this review paper.
• Figure 3: MDP for RL loophaydari2020deep.
• Figure 4: Actor Critic methodhaydari2020deep.
• Figure 5: Model-based RL