Research on Multi-target Automatic Localization of Inspection Robot Based on Adaptive Spatial Information in Unknown Environment

Abstract

The current multi-target search and localization algorithms for robots rely on a pre-established multi-target automatic positioning model. When an unpredictable multi-target navigation plan is encountered during the robot inspection process, it is easy to cause misjudgment and identification. For multi-target positioning and tracking in the background environment, high-confidence models are extracted from unstable information sources as observations. Considering the timeliness of the search method, we propose a perceptual adaptive target search method using a sub-area search strategy. The adaptive search method designs a search track for the robot online, and divides the search area in the field of view into several sub-areas. When there are no targets in the sight range, you can use the sub-area search strategy to improve the search efficiency and ensure the effective implementation of target search and positioning tasks. Compared with other optimized trajectories designed off-line, it can ensure multi-target search purpose and avoid missing key information. Considering the threat of uncertain obstacles in an unknown environment, according to the principle of timeliness, an improved dynamic window method is applied to avoid obstacles. Finally, simulations verify the adaptive search ability of the above method. This algorithm can stably track multiple targets in complex scenes, effectively distinguish different targets, and has a good processing effect on target occlusion and background interference.