Actuation tuning and sensor calibration are essential procedures required to obtain expected and desired behavior from an optionally-manned platform. Unfortunately, these tasks often require custom and brittle procedures which are time consuming and difficult to conduct outside of a controlled environment with special-purpose test equipment. Furthermore, many of these procedures require routine re-tuning and re-calibration by dedicated personnel due to disturbances in sensor mounting locations, general wear on vehicle components, environmental variations, and replacement of parts due to failure or damage. To streamline the process of tuning and calibration, we propose a method of automatically determining actuation parameters, sensor calibration, and vehicle models. This effort will leverage our prior work on the ONR Code 30 Autonomy program, and in Phase I will focus on further developing the capability of automatic vehicle model parameter discovery, extending algorithms to support sensor self-calibration, and evaluating performance of the prototype in a benchtop environment with both simulated data and data from an unmanned ground vehicle. Upon demonstration of viability, in Phase II we plan to transition and optimize our process to run on-line, allowing for real-time adaptation of the models to changing vehicle and terrain conditions.
