Biologically-Inspired Self-Tuning Neuromorphic PID Control for Real-Time Trajectory Tracking of a UAS

Abstract

This study presents a control framework for Unmanned Aerial Systems (UASs) that leverages the features of Spiking Neural Networks (SNNs). Towards this goal, we developed two neuromorphic controllers using Nengo - a prototyping environment for the deployment of biologically-plausible Neural Networks (NN). Initially, we implemented a neuromorphic Proportional-Integral-Derivative controller, which was subsequently enhanced with self-tuning capabilities. The proposed self-tuning controller is capable of dynamically adjusting its gains using a gradient descent optimization algorithm. The proposed controllers were assessed through a series of real-time experimental tests, where a DJI Tello UAS was tasked with performing trajectory tracking tasks.