Scholarlink

This work presents a biologically plausible approach to reinforcement learning control of benchmark cart-pole system with continuous state and discrete actions. The proposed solution utilizes brain-inspired spiking neural networks (SNN), constructed with the aid of the NEST simulator in Python. To update the SNN connection weights, the Temporal Difference TD(0) algorithm is combined with Spike Timing Dependent Plasticity (STDP), reflecting principles of reinforcement learning. It is shown that SNN actor-critic architecture solves the task several times faster than classical neural networks.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Reinforcement Learning Control of Cart Pole System with Spike Timing Neural Network Actor-Critic Architecture

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Borislav Markov,
  • Petia Koprinkova-Hristova

摘要

This work presents a biologically plausible approach to reinforcement learning control of benchmark cart-pole system with continuous state and discrete actions. The proposed solution utilizes brain-inspired spiking neural networks (SNN), constructed with the aid of the NEST simulator in Python. To update the SNN connection weights, the Temporal Difference TD(0) algorithm is combined with Spike Timing Dependent Plasticity (STDP), reflecting principles of reinforcement learning. It is shown that SNN actor-critic architecture solves the task several times faster than classical neural networks.