Diverse traversability mapping using hybrid vision system for lunar rover

Autonomous functions including mapping and path planning are very important to ensure efficient exploration and safe navigation of unmanned rover. Conventional navigation schemes based on equipments such as GPS and magnetic sensors have been proven to be ineffective on lunar surface, while sophisticated dead reckoning approach lacks accuracy in loose soil due to slippage. In this paper, we propose a hybrid vision system (HVS) which consists of one stereo and one omnidirectional vision sensor, attempting to integrate the advantages of diverse vision systems. The hierarchical mapping reconstructs environment through HVS in different resolutions and structures, named qualitative, rough and detailed map. These maps are employed for path planning which navigates the rover towards specified destinations autonomously, avoiding obstacles in its way. The HVS aims at providing more efficient and safe navigation. Simulations and experiments demonstrate the functionality and advantage of our approach.