SNN-Based Joint-Space Control of the Serv-Arm Robot on Low-Cost Embedded Hardware

Resumo

Resumo

This work introduces an SNN-based joint-space controller for a four-degree-of-freedom (DoF) servo-arm robotic manipulator, implemented via a spiking neural network (SNN) on low-cost embedded hardware. Unlike traditional methods based on inverse kinematics or Cartesian-space control, the proposed architecture operates directly on servo joint angles, receiving current and target configurations to predict the next joint-space action. A hybrid dataset with over 500,000 samples was generated by expanding joint-angle logs into a larger synthetic set to provide coverage-oriented sampling within the robot’s operational joint limits, including additional emphasis on mechanically challenging configurations. The SNN was trained via surrogate gradients (snnTorch), and a systematic grid search was used to optimize the hidden-layer size, learning rate, and membrane decay factor. The final model, featuring a 128-neuron leaky integrate-and-fire (LIF) hidden layer, achieved a mean absolute error (MAE) of 11.06° on the validation set and generated smooth, reproducible trajectories during real-robot execution under a low-frequency open-loop supervisory control setting. Task-level validation was performed through repeated real-platform pick-and-place executions, where the ANN baseline achieved lower final joint-space error and better waypoint proximity coverage than the SNN controller. Additionally, this study includes an empirical energy comparison between the SNN controller and an artificial neural network (ANN) baseline on a Raspberry Pi under identical motion routines. The results indicate that the SNN consumed slightly more average power than the ANN baseline on the Raspberry Pi due to temporal simulation overhead associated with software-based spike dynamics execution on conventional CPUs. Thus, the present implementation should be interpreted as SNN-based control executed on standard embedded CPU hardware rather than as control executed on dedicated neuromorphic hardware.

palavras-chave

Palavras-chave

spiking neural networksneuromorphic controljoint space controlrobot manipulatorslow-cost embedded hardwaresynthetic data generationleaky integrate-and-fire neuronsenergy consumptionRaspberry Pigrid search hyperparameter tuning

links

Links e arquivos