AgentSkillsCN

sumo-rl

利用 sumo-rl 进行强化学习交通信号控制，包括 SumoEnvironment 配置、Gymnasium/PettingZoo API、观测/动作/奖励设计，以及训练工作流。适用于 RL 信号控制任务，或在被要求使用 SUMO-RL 时使用。

SKILL.md
--- frontmatter
name: sumo-rl
description: Reinforcement learning traffic signal control with sumo-rl, including SumoEnvironment setup, Gymnasium/PettingZoo APIs, observation/action/reward design, and training workflows. Use for RL signal control tasks or when SUMO-RL is requested.

Sumo RL

Overview

Use this skill to design and run RL traffic signal control experiments with sumo-rl.

Skill Routing

  • •Use sumo-env for installing SUMO or Python dependencies.
  • •Use sumo-core to build networks, routes, and traffic light definitions.
  • •Use sumo-mcp for automated RL training workflows or tool execution.
  • •Use sumo-output for output analysis outside RL logs.

RL Workflow

  1. •Prepare a network with traffic lights and a route file.
  2. •Choose single-agent (Gymnasium) or multi-agent (PettingZoo) API.
  3. •Select observation and reward functions.
  4. •Train with your RL library or built-in examples.

Requirements

  • •SUMO_HOME set and SUMO executables available.
  • •Traffic lights must exist in the network for signal control.

References

  • •Installation: references/rl-install.md
  • •Observations, actions, rewards: references/rl-mdp.md
  • •Gymnasium and PettingZoo APIs: references/rl-env-api.md