Abstract:

The dynamic obstacles during the inspection process of intelligent robots result in discontinuous accessibility conditions for the inspection path， leading to abrupt changes in the feasible domain of the path and difficulty in obtaining feasible solutions for reconstructing the path， thereby reducing the smoothness of turning in the inspection path planning. To this end， an intelligent robot adaptive inspection path grid map neighborhood search algorithm is proposed. Adopting the grid map method to model the inspection environment， and defining the feasible path domain through the boundary condition processing mechanism. Introducing a turning cost function to improve the A^* algorithm， using a grid map feasible region as the inspection environment model， to generate a preliminary smooth inspection path for turns. Aiming at the failure problem caused by path mutations in dynamic environments， a feasible region dynamic reconstruction method for inspection paths based on adaptive large-scale neighborhood search is proposed. The destruction operator is used to remove some inspection tasks and generate incomplete paths， and the repair operator is introduced to reorganize inspection tasks. Combined with the variable neighborhood descent search strategy， the solution space is deeply explored. The local optimization of the preliminary global inspection path is achieved through the adaptive neighborhood operator scoring and selection mechanism. The experimental results show that the algorithm can successfully avoid dynamic and static obstacles in grid maps and achieve intelligent robot inspection path planning. The inspection path length is relatively short， and the path optimization degree， turning smoothness coefficient， global search ability， and environmental adaptability scores are 95.24%， 0.917， 0.764， and 91.45， respectively.

Key words: raster map, improve the A^* algorithm, heuristic function, cost of turning, neighborhood search, variable neighborhood descent