Spiking Neural Networks for Continuous Control via End-to-End Model-Based Learning
By: Justus Huebotter , Pablo Lanillos , Marcel van Gerven and more
Potential Business Impact:
Robots learn to move arms smoothly and accurately.
Despite recent progress in training spiking neural networks (SNNs) for classification, their application to continuous motor control remains limited. Here, we demonstrate that fully spiking architectures can be trained end-to-end to control robotic arms with multiple degrees of freedom in continuous environments. Our predictive-control framework combines Leaky Integrate-and-Fire dynamics with surrogate gradients, jointly optimizing a forward model for dynamics prediction and a policy network for goal-directed action. We evaluate this approach on both a planar 2D reaching task and a simulated 6-DOF Franka Emika Panda robot. Results show that SNNs can achieve stable training and accurate torque control, establishing their viability for high-dimensional motor tasks. An extensive ablation study highlights the role of initialization, learnable time constants, and regularization in shaping training dynamics. We conclude that while stable and effective control can be achieved, recurrent spiking networks remain highly sensitive to hyperparameter settings, underscoring the importance of principled design choices.
Similar Papers
Fully Spiking Actor-Critic Neural Network for Robotic Manipulation
Robotics
Robots learn to grab things faster, using less power.
Population-Coded Spiking Neural Networks for High-Dimensional Robotic Control
Robotics
Robots use less power while still moving well.
Spiking Neural Networks: a theoretical framework for Universal Approximation and training
Optimization and Control
Makes brain-like computers learn and work better.