摘要

Inspired by the excellent navigational behavior of animals, this paper presents a bionic autonomous navigation (bio-navi) system by using a stereo camera and a polarization navigation sensor (POLNS). A topological graph consisting of nodes and edges forms the core of the system, which is similar to the simultaneous localization and mapping (SLAM). Like most of the SLAM approaches, in the bio-navi system, a node contains pose (position and attitude) and visual template information while an edge represents a constraint between two nodes. The difference is that the POLNS is introduced into the system. The yaw angle obtained from POLNS is attached to a node as a new property and also forms a new edge which restricts the direction of the related node. Besides, an absolute place constraint edge is added to the current node when recognizing a prior known place. To improve the accuracy and robustness of the place recognition, a POLNS-assist loop closure detection method is presented and to correct the pose error of constructed nodes, an improved graph optimization algorithm is derived in detail which is able to make full use of all the constraints. To demonstrate the performance of the system, an outdoor vehicle test has been done, and the results show its effectiveness and feasibility.