Event Location: NSH 1507

Abstract: Advances in hardware design have made wheeled mobile robot(WMRs) exceptionally mobile. To fully exploit this mobility, WMR planning, control,and estimation systems require motion models that are fast and accurate. Much of the published theory on WMR modeling is limited to 2D or kinematics, but 3D dynamic (or force-driven) models are required when traversing challenging terrain, executing aggressive maneuvers, and manipulating heavy payloads. This thesis addresses the formulation and calibration of such models.

First, we present novel WMR model formulations that are high fidelity, general, modular, and fast. Our “stabilized DAE” kinematics formulation enables constrained, drift-free motion prediction on rough terrain. We also enhance the kinematics to predict nonzero slip in a principled way based on gravitational, inertial, and dissipative forces. Our constrained dynamics formulation permits large integration time steps without compromising stability. In addition, it can enforce realistic, nonlinear models of wheel-terrain interaction (e.g. terramechanics-based, empirical) using a novel “force-balance optimization” technique. Any articulated WMR design can be modeled with just a few lines of code, and simulated 1K-10K times faster than real-time on an ordinary PC. Simulation tests show our model formulations to be more functional, stable, and efficient than common alternatives.

Next, we present a novel Integrated Prediction Error Minimization (IPEM) method of calibrating these models that is general, convenient, online, and evaluative. Typically, system dynamics are calibrated by minimizing the error of instantaneous output predictions, but IPEM instead forms predictions by integrating the system dynamics over an interval. Benefits include reduced sensing requirements, better observability, and accuracy over a longer horizon. Experimental results on multiple platforms and terrain types show that parameter estimates converge quickly during online calibration, and uncertainty is well characterized. They also show significant improvement in predictive accuracy when models account for nonzero wheel slip and 3D articulations.

Committee:Alonzo Kelly, Chair

David Wettergreen

George Kantor

Karl Iagnemma, Massachusetts Institute of Technology