Automated locomotion parameter tuning for an Anguilliform-inspired robot

Effectiveness of the locomotion of an Anguilliform-inspired robot depends upon the selection of various controller parameters such as amplitude, frequency, offset and low level controller gains. Manual tuning of the aforementioned parameters can be cumbersome. Automated controller parameter tuning requires repetitive experimental tests with non-optimal parameters leading to a rapid wear and tear of the robot. Use of dynamics simulator instead of a robot may alleviate this issue to some extent. However, reality gap may exist between a physical system and corresponding simulator. This paper reports an optimization based approach for determining environment parameters (added mass, added inertia and drag) from physical experiments followed by an optimization based automated parameter tuning utilizing the determined environment parameters via developed dynamics simulator. We report numerical simulations of waypoint following on different test paths using the identified environment parameters and optimized controller parameters determined using the developed approach. We found a reduction of up to 36% in travel time in our experiments. The developed approach can lead to a match between dynamics simulation and physical experiments and thereby can help in automated parameter tuning. In future, we would like to improve the performance of the optimization and to incorporate the sensing noise and environmental disturbance like current into the dynamics simulator.