To solve the problems of low exploration efficiency and imprecise depth in object goal navigation,this article constructs a framework to solve object goal navigation.Depth map edge processing and map error correction mechan-isms were introduced in the semantic map construction module;a coverage maximization algorithm was introduced in the exploration module;alternative point mechanisms was introduced in the path planning module.This article conduc-ted experiments in a 3D simulation environment.The experimental results show that the new solution proposed in this article significantly improves the performance of object goal navigation.In addition,the method proposed in this article was successfully applied to quadruped robots,thereby verifying its generalization in real-world scenarios.
维普
万方数据