A Unified Perturbative Dynamics Approach to Online Vehicle Model Identification

Abstract

The motions of wheeled mobile robots are largely governed by contact forces between the wheels and the terrain. Inasmuch as future wheel-terrain interactions are unpredictable and unobservable, high performance autonomous vehicles must ultimately learn the terrain by feel and extrapolate, just as humans do. We present an approach to the automatic calibration of dynamic models of arbitrary wheeled mobile robots on arbitrary terrain. Inputs beyond our control (disturbances) are assumed to be responsible for observed differences between what the vehicle was initially predicted to do and what it was subsequently observed to do. In departure from much previous work, and in order to directly support adaptive and predictive controllers, we concentrate on the problem of predicting candidate trajectories rather than measuring the current slip. The approach linearizes the nominal vehicle model and then calibrates the perturbative dynamics to explain the observed prediction residuals. Both systematic and stochastic disturbances are used, and we model these disturbances as functions over the terrain, the velocities, and the applied inertial and gravitational forces. In this way, we produce a model which can be used to predict behavior across all of state space for arbitrary terrain geometry. Results demonstrate that the approach converges quickly and produces marked improvements in the prediction of trajectories for multiple vehicle classes throughout the performance envelope of the platform, including during aggressive maneuvering.

Slip-aware Model Predictive optimal control for Path following

“…We follow the approach described in [7] to estimate the components of wheel slip. In that approach, the system dynamics is linearized to capture, to the first order, the error in the predicted 6 DOF pose of the WMR given a disturbance to the input (which is assumed to be caused by wheel slip).…”

Experimental analysis of models for trajectory generation on tracked vehicles

“…wheeled and tracked vehicles and provides what is perhaps the only complete dynamic model for a turning tracked vehicle, though it is for quasi-static motion and may not be applicable to small platforms. It is important to note that some of these works explicitly seek to perform online estimation of the slip models by using an IMU [4], differential-GPS [7], or visual odometry [12].…”

“…We treat the effects of unmodeled vehicle dynamics, including tread-terrain interactions, by incorporating corrections to the effective velocity of the platform. These corrections are established by experimentally fitting relevant models from the literature [8], [12] to a series of maneuvers performed over a range of speeds and turning rates. By comparing the difference between the predicted and actual final position at the end of each motion, we show the value of using these corrections to help remove the systematic bias from the trajectory integration.…”

“…The authors of [8] point out that this is equivalent to considering an ideal differential drive model with an effectively larger wheel base. Lastly, in the General Kinematic Slip model (our modification of the model presented in [12]), we consider that slip conditions may lead to velocity changes in both the longitudinal and lateral directions. Following the authors of [12], we allow the correction to be velocity-dependent by defining:…”

“…However, since we lack the sensor capabilities to accurately measure platform velocities in the field, we settle instead for studying the perturbative effect of the parameters on the integrated trajectory in the spirit of [14]. As presented in [12], this could be done by looking at a larger motion where we can get accurate relative pose information over extended time-slices (such as by visual odometry or a tactical-grade INS) and then fitting parameters to minimize the error between expected final pose and actual final pose. For this effort, we chose to use a motion capture system for the sake of establishing an accurate baseline to compare these more field-appropriate techniques to in the future.…”