Multi-Vision Sensor Fusion Localization and Navigation of Robot in Unstructured Environment

Robots that have the ability to work in complex natural environments have broad application prospects. Focusing on the lack of autonomous navigation ability of the robot in the wild environment, this paper respectively proposes the improved localization and trajectory tracking methods. The multi-vision sensor fusion localization method is based on dual-factor graph optimization, which realizes the rapid processing of asynchronous sensor information. The dual-factor graph optimization can improve the localization accuracy and fault tolerance of the system under outdoor environments. Meanwhile, the multi-fine granularity MPC-based trajectory tracking method for unstructured terrain. Under the large-scale global path, a refined map is built around the robot. And the elevation difference is introduced to impose new constraints at the omni-directional velocity. The proposed methods in this study are implemented in actual environment. Experiments and simulations in representative terrain environments have verified the effectiveness of the improved methods.