Soft Robotics with Quantum-Driven Electronic Neural Networks

Abstract

Although this field attracts lots of attention, conventional control mechanisms of Soft Robotics are still restricted in real-time decision making, learning efficiency, and energy consumption. This research further strengthens soft robotic intelligence to present a novel Quantum Driven Electronic Neural Network (QD-ENN) framework based on Design of Reservoir Computing (QRC) to be used for development of Brain of the Offspring (BO) and contextual entangled processing (CEPT) nodes. Quantum superposition and entanglement make it intrinsically superior to sensorimotor learning, low power computation, and rapid adaptation of the sensorimotor interaction in an unstructured environment. Compared with classical deep learning methods that require huge quantities of training and computations to learn, the proposed system solves real-time control problem and changes morphologies of soft actuators dynamically using quantum inspired neural plasticity. Based on the design of the architecture, which is implemented for neuromorphic processing using memristor electronic synapses and based on quantum circuits to help with reinforcement learning, the architecture designed employs quantum circuits and memristor electronic synapses. Experimental evaluations also demonstrate excellent speed up in terms of learning speed, decision accuracy and energy efficiency compared to the traditional AI-driven soft robotic controllers. Based on this work, future research on quantum neuromorphic architectures in robotics follows by building semiconductor hardware towards self-learning robotics of exceptionally dynamical and unpredictable nature.