Reproduce “Adaptive Robotic Arm Control with a Spiking Recurrent Neural Network on a Digital Accelerator” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Alejandro Linares-Barranco, Luciano Prono, Robert Lengenstein, Giacomo Indiveri, Charlotte Frenkel

Researchers successfully implemented the open-source ReckOn spiking recurrent neural network accelerator on a Xilinx Multiprocessor System on Chip (MPSoC) to enable adaptive robotic arm control. The system achieved 83.3% test accuracy in classifying a 1-kg weight attached to a robotic arm and demonstrated a peak event processing rate of 3.8 Million events per second.

Key insights

  • Porting a dedicated neuromorphic ASIC design to a flexible, programmable MPSoC significantly enhances its accessibility and enables broader application in embedded systems and research.
  • The integration of a hardware-accelerated spiking neural network with a high-level software interface (Python on Jupyter) facilitates rapid prototyping and experimentation for online learning at the edge.
  • Spike-based sensory inputs derived from robotic motor controllers can be effectively processed by on-chip RSNNs to learn complex, real-time adaptive control tasks in physical robotic systems.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.