Autonomous Traversal of Rough Terrain Using Behavioural Cloning

Behavioural cloning is a method for acquiring skills by building generalised behaviours based on the observations of human performance. In this work, we examine the application of behavioural cloning to autonomous navigation of a robot in an unstructured environment. In particular, we examine the traversal of the random stepfields, introduced in the Robocup Rescue Robot League competition, by a tracked vehicle. One critical issue in behavioural cloning is representing the state of the environment in a manner amenable to machine learning. Our representation and behavioural cloning technique was evaluated by implementing it on our CASTER advanced mobility robot and training it to traverse a sequence of two NIST-specification stepfields. Despite being derived from only ten training runs, the cloned behaviour was able to successfully traverse the stepfield unaided 40% of the time and required only minor intervention the remaining 60% of the time.