Spiking Neural Network-based Control Applied to an Underactuated System

This paper presents the problem of stabilizing an underactuated system, the Ball and Plate platform, by means of a neuromorphic-inspired control. The proposed architecture makes use of Spiking Neural Networks (SNNs) in combination with the Neural Engineering Framework (NEF) to accomplish regularization and trajectory tracking. Simulation results for both the proposed SNN controller and its conventional (non-neural) counterpart are presented and qualitatively compared. The results show that the SNN controller effectively tracks a desired trajectory with minimal tracking errors. The practical considerations for implementing a fully neuromorphic real-time application of the proposed approach are discussed as well.