Using kinesthetic input to overcome obstacles with snake robots

Snake robots have enormous potential to thread through tightly packed spaces and relay knowledge to search and rescue workers which is currently unattainable during the first hours of rescue operations. However, the existing approaches to snake robot locomotion in three dimensions is primarily limited to cyclic gaits, which lose effectiveness as the ratio of obstacle size to robot size or the irregularity of the environment increase. To this end, this work investigates a kinesthetic input approach to developing joint angle trajectories for overcoming these obstacles for which gaits are inadequate. The second contribution of this paper is the presentation and validation of a method to simplify these trajectories so that they can be easily stored, parameterized, and adjusted. Finally, we demonstrate that a simple sensor deviation filtering and thresholding approach can be used to quickly detect failure when overcoming an obstacle.