Design of Autonomous Navigation System for Quadrotor in Subway Tunnel

Aiming at the subway tunnel environment with weak light and no GPS signal, we build a hardware platform of quadrotor based on lidar, airborne upper computer and bottom flight control board. Combined with HectorSLAM technology, we also design a route planning algorithm based on the improved Rapidly-exploring Random Tree (RRT) and Artificial Potential Field Method (APF), in order to realize intelligent inspection of the subway tunnel. Firstly, for the path planning of large scene map, an improved algorithm Rapidly-exploring Random Tree with Regression Filtering Mechanism and Trunk Constraint (ReT-RRT
^*
) is proposed to reduce the scope of invalid search space, avoid excessive search of explored areas and effectively save path planning time. Secondly, we use cubic B-spline to optimize the planned route generated by ReT-RRT
^*
and improve tracking accuracy. Thirdly, we introduce `route point' gravity to improve traditional APF, in order to realize online obstacle avoidance combined with the planned route. Finally, we carry out experiments to verify the effectiveness of the improved methods and the feasibility of the design of autonomous navigation system.