Model-Based External Wrench Estimation for Underwater Robots

Abstract

Similarly to aerial drones, small-scale underwater robots are prone to external wrenches resulting from disturbances such as water currents or collisions. Estimating the external wrench acting on an underwater robot is challenging due to non-linear hydrodynamic effects and the bottleneck of being limited to onboard sensing.We build on a model-based approach for aerial wrench estimation and extend it to the underwater domain. Various modifications are applied, such as capturing hydrodynamic effects, and new sensory information is integrated, for example, via Doppler velocity log (DVL).We evaluate the performance of the proposed approach through a series of experiments. Moreover, we assess the effect of fusing various sensor configurations and their respective influence on the wrench estimate, including low-cost vs. high-end IMU and DVL. Our adapted approach from the aerial domain delivers good results in estimating external wrenches on underwater robots. While the IMU quality is found to be less important, considering the underwater domain-specific damping terms is critical.