Abstract

In the realm of intelligent transportation, autonomous driving relies heavily on robust path planning. This study introduces a novel Deep Reinforcement Learning algorithm for Path Planning (DRL-PP) to address limitations in current methodologies. To navigate complex environments, DRL-PP is engineered to determine optimal actions while minimizing overfitting. The algorithm employs neural networks to identify advantageous state-specific actions, constructing an optimized trajectory from origin to destination. A critical innovation lies in the enhanced reward function, which integrates objective-specific data to dynamically calibrate reward metrics. This refinement significantly augments decision-making efficiency. Empirical evaluations demonstrate that DRL-PP stabilizes reward accumulation and minimizes exploratory iterations. Comparative analysis confirms that the proposed algorithm consistently outperforms benchmark models in navigational efficacy, offering a robust solution for the evolutionary advancement of autonomous vehicle technology.