No abstract available.

We are developing a mobile robot capable of emulating general 6-degree-of-freedom spacecraft relative motion. The omni-directional base uses a trio of active split offset castor drive modules to provide smooth, holonomic, precise control of its motion. Encoders measure the rotations of the six wheels and the three castor pivots. We present a generic odometric algorithm using a least squares framework which is applicable to vehicles with two or more castors and apply it to our unique vehicle configuration. As the accuracy of odometry algorithms depends on the accuracy to which the model parameters are known, a method to perform calibration on the physical robot is needed. We present a geometric calibration method based solely on internal sensor measurements. We present a range of simulation results comparing our odometry results to other algorithms under various systematic and non-systematic errors. We evaluate the ability of our calibration method to accurately determine the true values of our system parameters. The odometry algorithm was also implemented and tested in hardware on our robotic platform. The results presented in the paper validate the calibration and odometry algorithms in both simulation and hardware.

This paper presents an improved Adaptive-Reinforcement Learning Control methodology for the problem of unmanned air vehicle morphing control. The reinforcement learning morphing control function that learns the optimal shape change policy is integrated with an adaptive dynamic inversion control trajectory tracking function. An episodic unsupervised learning simulation using the Q-learning method is developed to replace an earlier and less accurate Actor-Critic algorithm. Sequential Function Approximation, a Galerkin-based scattered data approximation scheme, replaces a K-Nearest Neighbors (KNN) method and is used to generalize the learning from previously experienced quantized states and actions to the continuous state-action space, all of which may not have been experienced before. The improved method showed smaller errors and improved learning of the optimal shape compared to the KNN.