Embodied Neuromorphic Control Applied on a 7-DOF Robotic Manipulator

TL;DR

Spiking Neural Network is used to leverage the spatiotemporal continuity of the motion data to improve control accuracy, and eliminate manual parameters tuning in embodied neuromorphic control of 7 degree-of-freedom robotic manipulators.

Abstract

The development of artificial intelligence towards real-time interaction with the environment is a key aspect of embodied intelligence and robotics. Inverse dynamics is a fundamental robotics problem, which maps from joint space to torque space of robotic systems. Traditional methods for solving it rely on direct physical modeling of robots which is difficult or even impossible due to nonlinearity and external disturbance. Recently, data-based model-learning algorithms are adopted to address this issue. However, they often require manual parameter tuning and high computational costs. Neuromorphic computing is inherently suitable to process spatiotemporal features in robot motion control at extremely low costs. However, current research is still in its infancy: existing works control only low-degree-of-freedom systems and lack performance quantification and comparison. In this paper, we propose a neuromorphic control framework to control 7 degree-of-freedom robotic manipulators. We use Spiking Neural Network to leverage the spatiotemporal continuity of the motion data to improve control accuracy, and eliminate manual parameters tuning. We validated the algorithm on two robotic platforms, which reduces torque prediction error by at least 60% and performs a target position tracking task successfully. This work advances embodied neuromorphic control by one step forward from proof of concept to applications in complex real-world tasks.

Figures (11)

• Figure 1: Flowchart of Trajectory Tracking Application. The algorithm predicts the torque based on the desired trajectory (joint positions, velocities, and accelerations), and the feedback controller corrects the predicted torque. Sensors read the actual torque applied to the robotic arm and feed it to the algorithm as a learning signal.
• Figure 2: Single Neuron Simulation. The post-neuron receives and integrates the current transmitted from the pre-neuron, with its membrane potential gradually accumulating over simulated time steps until it fires a spike and returns to the resting potential.
• Figure 3: Overall Control Framework. The control framework adopts closed-loop control, where sensors on the robotic arm output the current state and the actual torque applied to the arm, ensuring that the robotic arm accurately reaches the target state and tracks the desired trajectory.
• Figure 4: Spike Encoding
• Figure 5: Neural Network