Unbounded Motion Optimization by Developmental Learning

An algorithm is presented for autonomous motion development with unbounded waveform resolution. Rather than a single optimization in a very large space, memory is built to support incremental improvements; therefore, complexity is balanced by experience. Analogously, human development manages complexity by limiting it during initial learning stages. Motions are represented by cubic spline interpolation; therefore, the development technique applies broadly to function optimization. Adding a node to the splines allows all previous memory samples to transfer to the higher dimension space exactly. The memory-based model, which is a locally weighted regression (LWR), predicts the expected outcome for a motion and provides gradient information for optimizing the motion. Results are compared against bootstrapping a direct optimization (DO) on a mathematical problem. Additionally, the method has been implemented to learn voltage profiles with the lowest peak current for starting a motor. This method shows practical accuracy and scalability.