Memory of Motion for Warm-starting Trajectory Optimization
share
Trajectory optimization for motion planning requires a good initial guess to obtain good performance. In our proposed approach, we build a memory of motion based on a database of robot paths to provide a good initial guess online. The memory of motion relies on function approximators and dimensionality reduction techniques to learn the mapping between the task and the robot paths. Three function approximators are compared: k-Nearest Neighbor, Gaussian Process Regression, and Bayesian Gaussian Mixture Regression. In addition, we show that the usage of the memory of motion can be improved by using an ensemble method, and that the memory can also be used as a metric to choose between several possible goals. We demonstrate the proposed approach with the motion planning on a dual-arm PR2 robot.
READ FULL TEXT