This paper presents a novel Spiking Neural Network (SNN) Controller for high-level kinematic control of a mobile robot. Such a controller is equipped with a Pure Pursuit (PP) path planning logic to ensure smooth movement across predetermined random waypoints and, in kinematic simulations across 100 random scenarios, it is benchmarked against three alternatives: the PI expert controller, a structurally similar Artificial Neural Network (ANN) controller (all embedded with such PP logic), and the classical Pure-pursuit controller. Simulation results show comparable performance in tracking (average error 0.33 ∼m on a 40 -by- 40 meter map). Crucially, due to its sparse event-driven processing, the SNN achieves lower (approximately halved) estimated computational energy consumption than the dense ANN counterpart, while preserving equivalent tracking performance, highlighting its primary advantage for mobile, possibly resource-constrained, robot platforms. As a preliminary, empirical robustness analysis, our approach was further validated in an additional, highfidelity simulation of a Stretch RE-2 robot in Simscape/Simulink environment with added measurement and control noise, maintaining proper tracking and achieving a Mean Absolute Error of 0.58 m on a new unforeseen trajectory. Finally, successful runs stemming from the deployment of SNN controller on the physical Stretch RE-2 robot further confirm the validity and viability of the proposed controller for real-world applications.
Spiking Neural Network Control for Mobile Robot Path Following / Troconis, L.G., Vella, F., Chen, Q., Freddi, A., Genova, O., Monteriu', A.. - (2026), pp. 601-606. (34th Mediterranean Conference on Control and Automation, MED 2026 ita 2026) [10.1109/MED70602.2026.11598557].
Spiking Neural Network Control for Mobile Robot Path Following
Troconis L. G.;Vella F.;Freddi Alessandro;Genova O.;Monteriu' A.
2026-01-01
Abstract
This paper presents a novel Spiking Neural Network (SNN) Controller for high-level kinematic control of a mobile robot. Such a controller is equipped with a Pure Pursuit (PP) path planning logic to ensure smooth movement across predetermined random waypoints and, in kinematic simulations across 100 random scenarios, it is benchmarked against three alternatives: the PI expert controller, a structurally similar Artificial Neural Network (ANN) controller (all embedded with such PP logic), and the classical Pure-pursuit controller. Simulation results show comparable performance in tracking (average error 0.33 ∼m on a 40 -by- 40 meter map). Crucially, due to its sparse event-driven processing, the SNN achieves lower (approximately halved) estimated computational energy consumption than the dense ANN counterpart, while preserving equivalent tracking performance, highlighting its primary advantage for mobile, possibly resource-constrained, robot platforms. As a preliminary, empirical robustness analysis, our approach was further validated in an additional, highfidelity simulation of a Stretch RE-2 robot in Simscape/Simulink environment with added measurement and control noise, maintaining proper tracking and achieving a Mean Absolute Error of 0.58 m on a new unforeseen trajectory. Finally, successful runs stemming from the deployment of SNN controller on the physical Stretch RE-2 robot further confirm the validity and viability of the proposed controller for real-world applications.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


