Adaptive-Optimal Control for Reconfigurable Robots

Reconfigurable robots refer to the robotic systems which have the ability to transform their morphological structure to adapt to changes in their surroundings. A specialized control system based upon the dynamic characteristics is required to be modelled so that a vast range of operations can be performed by the reconfigurable robots. Reinforcement learning based technique is therefore considered as a suitable mechanism for carrying out adaptive optimal control for the robotic systems. In this paper, an attempt has been made to model the adaptive optimal control mechanism for a single-link manipulator: a reconfigurable robotic system. The system has been modelled for a dynamic environment. The loss and reward plots are obtained from the simulation of the robotic system by deploying reinforcement learning based techniques with the given controlled input