Synergetic Learning Control Paradigm for Redundant Robot to Enhance Error-Energy Index

In order to perform energetically efficient motion as in human control, so-called optimization-based approach is commonly used in both robotics and neuroscience. Such an optimization approach can provide optimal solution when the prior dynamics information of the manipulator and the environment is explicitly given. However, the environment, where the robot faces with in a real world rarely has such a situation. The dynamics conditions change by the contact situation or the hand load for the manipulation task. Simple computational paradigm to realize both adaptability and learning is essential to bridge the gap between learning and control process in redundancy. We verify a novel synergetic learning control paradigm in reaching task of redundant manipulator. The performance in handling different dynamics conditions is evaluated in dual criteria of error-energy (E-E) coupling without prior knowledge of the given environmental dynamics and with model-optimization-free approach. This paper aims at investigating the ability of phenomenological optimization with the proposed human-inspired learning control paradigm for environmental dynamics recognition and adaptation, which is different from conventional model optimization approach. E-E index is introduced to evaluate not only the tracking performance, but also the error reduction rate per the energy consumption.