A Vision and Behaviour Based Approach for Short-Range Autonomous Navigation of Planetary Rovers

Planetary rovers, an essential robotic tool in planetary expeditions, have been used for decades to study the Moon and the planet Mars. The problem with operating on Mars and the dark craters on the Moon is the long cycle communication delay, which makes it inefficient to remotely operate the rover from Earth. Taking the MER (Mars Exploration Rovers) mission as an example, the rovers are remotely controlled by the ground station to perform long-range navigation. The distance between waypoints, which allows the rover to travel autonomously in a short range, is reported to be limited to a short distance due to the computational complexity of the hazard avoidance algorithm. As a result, the rover can sometimes travel only a few meters per Martian day (sol). Increasing onboard autonomy of the rover can help to reduce human intervention and improve mission efficiency.

The aim of this project is to introduce and investigate a robust and fast onboard navigation system that allows the rover to travel autonomously and safely through a longer distance given a grand goal. We propose to use vision and behaviour based navigation algorithms based on 2D images collected by a single camera. The navigation algorithm relies on the
extracted information from image processing, such as features, motion vectors, focus of expansion (FOE), time-tocontact (TTC), and makes behavioural decisions to track targets and/or avoid obstacles. To use a simple sensor for the system has advantages of low power consumption and low computation and design complexity.