LPSim 是 AI Skill Hub 本期精选Agent工作流之一。综合评分 8.0 分，整体质量较高。我们强烈推荐将其纳入你的 AI 工具库，帮助提升工作效率。
LPSim 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排，将复杂的多步骤任务拆解为清晰的自动化流程，实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成，适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
LPSim 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排，将复杂的多步骤任务拆解为清晰的自动化流程，实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成，适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
# 克隆仓库 git clone https://github.com/Xuan-1998/LPSim cd LPSim # 查看安装说明 cat README.md # 按 README 完成环境依赖安装后即可使用
# 查看帮助 lpsim --help # 基本运行 lpsim [options] <input> # 详细使用说明请查阅文档 # https://github.com/Xuan-1998/LPSim
# lpsim 配置说明 # 查看配置选项 lpsim --config-example > config.yml # 常见配置项 # output_dir: ./output # log_level: info # workers: 4 # 环境变量（覆盖配置文件） export LPSIM_CONFIG="/path/to/config.yml"
LPSim is a GPU-accelerated, multi-GPU traffic microsimulator for HPC and AI research. It scales city-region networks with hundreds of thousands of nodes and millions of agents to run in minutes on commodity GPUs, and it is built to serve as a high-throughput environment for reinforcement-learning policies in mobility (dispatch, routing, pricing, vertiport siting).
The platform underpins our ICML 2026 paper on regime-aware deep RL for ride-hailing dispatch, our T-RC 2024 paper on multi-GPU traffic assignment, and ongoing work on urban air mobility (UAM) integration.
<img width="1200" alt="LPSim Bay Area" src="https://github.com/Xuan-1998/LPSim/assets/58761221/1c41f659-aee0-4887-99e0-39b0133154ce">
---
docker pull yibo123/lpsim:cuda12.4
git clone https://github.com/Xuan-1998/LPSim.git && cd LPSim docker run -it --rm --gpus all -v "$PWD":/lpsim -w /lpsim yibo123/lpsim:cuda12.4 bash
The Docker image bundles CUDA 12.4, Qt5, Boost, Pandana, and all build deps. This is the recommended path.
```bash
Tested on Ubuntu 22.04 / 24.04 with CUDA 12.3+, GCC 11+, Qt 5.15. Older versions (CUDA 9, GCC 6.4, Qt 5.9.5) still build but are no longer the supported path.
```bash
export PATH=/usr/local/cuda-12.3/bin:$PATH export LD_LIBRARY_PATH=/usr/local/cuda-12.3/lib64:$LD_LIBRARY_PATH
Pandana (CH routing backend) needs a shared library:
sh LivingCity/runAllTests.sh ```
LPSim exposes its simulator as a callable kernel-loop, which is convenient as a high-throughput environment for RL policies operating on city-scale fleets. The Python workflow under optimizer_*MSA_S3*.py and final_analysis_0930.py is the entry point used by our ICML 2026 work; treat those as reference integrations.
If you build a new RL/optimization pipeline on top of LPSim, please cite the relevant paper(s) below.
---
Edit LivingCity/command_line_options.ini:
[General]
NETWORK_PATH=data/networks/sf_bay_area/
USE_SP_ROUTING=true
USE_PREV_PATHS=true
NUM_PASSES=1
TIME_STEP=0.5
START_HR=5
END_HR=12| Key | Meaning |
|---|---|
NETWORK_PATH | Directory containing nodes.csv, edges.csv, demand. |
USE_SP_ROUTING | Use the CH (Pandana) shortest-path routing. Keep true. |
USE_PREV_PATHS | Reuse cached paths from a prior run. false on the first run. |
NUM_PASSES | Number of simulation passes per run. |
TIME_STEP | Simulation time step in seconds. 0.5 is the validated default. |
START_HR, END_HR | Simulation window, 24-hour clock. |
GUI, USE_CPU, USE_JOHNSON_ROUTING, LIMIT_NUM_PEOPLE, ADD_RANDOM_PEOPLE are legacy flags. Leave at defaults.
---
高性能的AI工作流，适合交通模拟和AI研究
AI Skill Hub 为第三方内容聚合平台，本页面信息基于公开数据整理，不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后，再部署至生产环境，并做好必要的安全评估。
✅ MIT 协议 — 最宽松的开源协议之一，可自由商用、修改、分发，仅需保留版权声明。
经综合评估，LPSim 在Agent工作流赛道中表现稳健，质量优秀。如果你已有明确的使用需求，可以直接上手体验；如果还在评估阶段，建议对比同类工具后再做决策。
| 原始名称 | LPSim |
| Topics | ai-for-sciencecudadeep-reinforcement-learninggpu-computinghigh-performance-computing |
| GitHub | https://github.com/Xuan-1998/LPSim |
| License | MIT |
| 语言 | Jupyter Notebook |
收录时间：2026-06-25 · 更新时间：2026-06-25 · License：MIT · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型，复制安装指令后粘贴到对应客户端