Path Planning for Orchard Mobile Robots Based on A*-Guided Improved Ant Colony Optimization

To address blind search, frequent deadlocks, slow convergence, and poor path quality in conventional ant colony algorithms for mobile robot path planning, this study proposes an A*-guided multi-strategy improved ant colony algorithm (MIACO). First, a global guidance path is generated via A*, and pheromones are non-uniformly initialized to enhance directionality and reduce blind exploration. Second, an ant death penalty mechanism applies negative feedback to lower pheromone concentration on deadlocked paths, reducing deadlock occurrence. Third, the heuristic function incorporates Euclidean distance to the goal, and an adaptive dynamic adjustment factor is introduced into the state transition probability to reduce invalid searches and accelerate convergence. Finally, the path node optimization strategy removes redundant nodes satisfying the direct-connectivity condition through geometric evaluation and obstacle detection, thereby reducing the number of path nodes and turns and improving the geometric quality of the planned path in the grid environment. Simulation results show that, in the simple grid environment, compared with ant colony optimization (ACO) and improved ant colony optimization (IACO), the path lengths obtained by MIACO are reduced by 11.35% and 3.94%, respectively; the average numbers of convergence iterations are reduced by 66.88% and 32.05%, respectively; and the average convergence times are reduced by 67.79% and 77.98%, respectively. In addition, compared with ACO, the average number of iterations and the average time required to achieve zero ant deaths are reduced by 36.47% and 49.38%, respectively. In the complex orchard simulation environment, compared with ACO and IACO, the path lengths obtained by MIACO are reduced by 33.15% and 2.18%, respectively; the average numbers of convergence iterations are reduced by 85.14% and 84.00%, respectively; and the average convergence times are reduced by 66.29% and 95.12%, respectively. The experimental results demonstrate that MIACO exhibits favorable path search and convergence performance in the two constructed grid environments.