Improved Ant Colony Algorithm Based on Reinforcement Learning for Robot Path Planning in Grid Environment

To address the issues of slow convergence, susceptibility to local optima and deadlock, and reduced population diversity in later stages of ant colony optimization (ACO), this paper proposes an improved ACO algorithm based on reinforcement learning (RBIACO). First, an exploration auxiliary factor is introduced to avoid deadlock and encourage ants to actively explore unvisited areas, thereby enhancing global search capability. Second, an improved pheromone increment formula is proposed, which assigns differentiated pheromone updates to nodes of different types within the same feasible path, alleviating the masking of key nodes caused by uniform updates in traditional methods. Furthermore, inspired by Q‐learning, a node value estimation mechanism is designed to evaluate the potential value of each node and is applied as a coefficient to the pheromone update process. Leveraging the search and self‐evolution features of ant colonies, this mechanism effectively improves overall algorithm performance. Finally, simulations on maps of varying scales were conducted, comparing RBIACO with basic ACO, particle swarm optimization (PSO), and IQBACO algorithms. The results demonstrate that RBIACO achieves significantly faster convergence, better global search ability, and higher search efficiency.