Robot Path Planning Based on Generative Learning Particle Swarm Optimization

Path planning refers to finding the optimal path from the starting point to the endpoint in a given environment, avoiding obstacles. To solve the problem of slow convergence speed in particle swarm optimization in path planning, this paper proposes a robot path planning based on generative learning Particle Swarm Optimization (LPSO). This algorithm constructs a generative double-adversarial network. In the first stage, the generator was used to analyze and process the initial map to obtain a foreground area with feasible paths. This area is used for heuristic search of particle swarms, reducing unnecessary exploration areas for particles throughout the state space, and quickly achieving path planning goals. In the second stage, the foreground region obtained in the first stage is used as the global optimal particle path for particle swarm optimization, and the particles are guided to move in the direction of high-density pheromones. Finally, the obstacle avoidance strategy enables the robot to avoid moving obstacles safely. In addition to being adapted to simple raster maps, this method also performs well in actual environment maps, showing superior generalization ability. To verify the effectiveness of the proposed algorithm, a series of simulation experiments are compared with the traditional PSO, ant colony algorithm (ACO), and improved algorithms, and the results show that under the same map environment, the LPSO algorithm has a faster convergence speed and shorter planning time and path length.