Rut detection and following for autonomous ground vehicles

When driving on off road terrains, experienced drivers utilize driving rules that have been developed through experience to handle difficult driving conditions such as traversing soft terrains, descending steep hills, and traversing water bodies among others. As robots move to challenging outdoor applications, it becomes imperative to adapt these off road driving rules to the robot control paradigms and hence improve the efficiency and safety of unmanned ground vehicles (UGVs).;One of the rules used by expert off road drivers is that when driving on soft terrains the vehicle should follow the prior vehicular transit. Rut following has the following benefits: (1) it prevents the ruts from serving as obstacles that can lead to undesirable vehicle vibrations and even vehicle instability at high speeds; (2) it improves vehicle safety on turns by utilizing the extra lateral force provided by the ruts to reduce lateral slippage and guide the vehicle through its path; (3) it improves the vehicle energy efficiency by reducing the energy wasted on compacting the ground; (4) it increases vehicle traction when traversing soft terrains such as mud, sand or snow.;This dissertation presents a set of field experiments to show the relevance of rut following for ground vehicles. Then, a laser based rut detection and following system is proposed so that autonomous ground vehicles can benefit from the application of this off road driving rule. The proposed system utilizes a path planning algorithm to aid in the rut detection process and an extended Kalman filter to recursively estimate the parameters of a local model of the rut being followed. Experimental evaluation on a Pioneer 3-AT robot shows that the proposed system is capable of detecting and following ruts in a variety of scenarios.