Repository navigation
修改选题所用配置引擎 #4043
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Closed
Closed
修改选题所用配置引擎 #4043
Changes from all commits
Commits
Show all changes
12 commits
Select commit
Hold shift + click to select a range
fa0864d
提交选题
lzk0804 c2d7324
Merge pull request #1 from lzk0804/main1
lzk0804 197efec
提交选题
lzk0804 b0c56d9
Merge pull request #2 from lzk0804/main1
lzk0804 9bb7d1f
提交选题
lzk0804 7e86dd6
Merge pull request #3 from lzk0804/main1
lzk0804 1dea8a6
提交选题
lzk0804 5694299
轮式小车可视化模型
lzk0804 faab519
Merge pull request #4 from lzk0804/main1
lzk0804 36ce28f
修改选题所用配置引擎
lzk0804 015b079
Merge pull request #5 from lzk0804/main1
lzk0804 5deb508
Merge branch 'main' into main
lzk0804 File filter
Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
There are no files selected for viewing
Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.
Oops, something went wrong.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,129 +1,139 @@ | ||
| # 简易轮式小车智能代理 | ||
|
|
||
| ## 项目简介 | ||
| 一个基于PyTorch神经网络与ROS的**简易轮式小车智能代理**,主要面向Carla自动驾驶模拟器。项目实现了从感知到控制的基本自动驾驶流程,模块化设计清晰,便于学习、扩展与二次开发。 | ||
|
|
||
| ## 功能特性 | ||
| - **环境感知**:处理相机图像,识别车道线、可行驶区域及动态障碍物。 | ||
| - **决策规划**:基于感知结果进行局部路径规划与动态避障决策。 | ||
| - **运动控制**:输出转向、油门及刹车指令,控制小车沿规划路径行驶。 | ||
| - **训练接口**:提供数据采集与模型训练接口,支持强化学习与模仿学习。 | ||
| - **ROS集成**:通过ROS节点进行模块间通信,支持传感器数据发布与控制指令订阅。 | ||
|
|
||
| ## 环境要求 | ||
| - **操作系统**:Ubuntu 20.04/22.04(推荐)或 Windows 10/11(WSL2) | ||
| - **Python**:3.7 - 3.12 | ||
| - **核心框架**:PyTorch >= 1.9.0(CPU/GPU版本均可) | ||
| - **模拟器**:Carla 0.9.14+ | ||
| - **可选依赖**:ROS Noetic 或 ROS2 Foxy(用于ROS通信) | ||
| - **基础包**:numpy, opencv-python, matplotlib | ||
|
|
||
| ## 项目结构 | ||
| ``` | ||
| src/simple_wheeled_agent/ | ||
| ├── main.py # 模块主入口 (非ROS启动) | ||
| ├── main.launch # ROS启动配置文件 | ||
| ├── requirements.txt | ||
| ├── config/ | ||
| │ └── default.yaml # 配置文件 | ||
| ├── docs/ # 模块文档 | ||
| ├── models/ # 预训练模型存放处 | ||
| ├── scripts/ # 工具脚本 | ||
| ├── src/ # 核心源代码 | ||
| │ ├── perception/ | ||
| │ │ ├── __init__.py | ||
| │ │ ├── image_processor.py # 图像处理与神经网络推理 | ||
| │ │ └── data_collector.py # 训练数据采集 | ||
| │ ├── planning/ | ||
| │ │ ├── __init__.py | ||
| │ │ └── local_planner.py # 局部路径规划器 | ||
| │ ├── control/ | ||
| │ │ ├── __init__.py | ||
| │ │ └── pid_controller.py # 运动控制器 | ||
| │ └── utils/ # 通用工具函数 | ||
| └── tests/ # 单元测试 | ||
| ``` | ||
|
|
||
| ## 快速开始 | ||
| 1. **安装依赖**: | ||
| ```bash | ||
| cd src/simple_wheeled_agent | ||
| pip install -r requirements.txt -i https://mirrors.aliyun.com/pypi/simple/ | ||
| ``` | ||
|
|
||
| 2. **启动Carla服务端** (在另一个终端): | ||
| ```bash | ||
| ./CarlaUE4.sh -quality-level=Low # 可根据硬件调整画质 | ||
| ``` | ||
|
|
||
| 3. **运行代理**: | ||
| ```bash | ||
| python main.py --config config/default.yaml | ||
| ``` | ||
| 程序将连接Carla,生成车辆,并开始自主驾驶。 | ||
|
|
||
| 4. **(可选) ROS模式**: | ||
| ```bash | ||
| roslaunch simple_wheeled_agent main.launch | ||
| ``` | ||
|
|
||
| ## 主要类和方法 | ||
| - **`PerceptionAgent` (感知模块)**: | ||
| - `process_image(raw_image)`: 处理原始图像,输出车道与障碍物信息。 | ||
| - `collect_training_data()`: 采集带标签的训练数据。 | ||
|
|
||
| - **`LocalPlanner` (规划模块)**: | ||
| - `compute_path(perception_result, goal)`: 根据感知结果和全局目标计算局部路径。 | ||
| - `dynamic_obstacle_avoidance()`: 动态避障算法。 | ||
|
|
||
| - **`VehicleController` (控制模块)**: | ||
| - `update_control(target_path, current_state)`: 根据目标路径和车辆状态计算控制量。 | ||
| - `get_control()`: 返回当前的控制指令(转向、油门、刹车)。 | ||
|
|
||
| ## 配置参数 | ||
| 关键参数可在 `config/default.yaml` 中调整: | ||
| ```yaml | ||
| carla: | ||
| host: localhost | ||
| port: 2000 | ||
| timeout: 10.0 | ||
| vehicle: | ||
| target_speed: 30.0 # 目标速度 (km/h) | ||
| max_steer_angle: 70.0 # 最大转向角 (度) | ||
| perception: | ||
| model_path: models/perception_model.pth | ||
| image_width: 640 | ||
| image_height: 480 | ||
| ``` | ||
|
|
||
| ## 数据输出 | ||
| 运行过程中会生成以下数据: | ||
| 1. **控制台输出**:车辆状态、控制指令、故障信息。 | ||
| 2. **日志文件** (位于 `logs/`): 运行日志与性能统计。 | ||
| 3. **数据记录** (训练模式下): 图像与对应控制指令,保存于 `data/train/`。 | ||
| 4. **可视化窗口**:实时显示处理后的图像与规划路径。 | ||
|
|
||
| ## 扩展建议 | ||
| - **更换模拟器**:参考现有Carla接口,为AirSim或MuJoCo实现对应的环境包装器。 | ||
| - **改进感知模型**:替换 `perception` 模块中的网络结构(如使用ResNet、Transformer)。 | ||
| - **尝试新算法**:在 `planning` 模块中实现强化学习(PPO、SAC)或更先进的规划算法。 | ||
| - **增加传感器**:集成激光雷达(Lidar)或雷达(Radar)数据进行多传感器融合。 | ||
|
|
||
| ## 常见问题 | ||
| 1. **Q: 无法连接到Carla服务器** | ||
| A: 请确认CarlaUE4正在运行,且端口未被占用。检查配置中的 `host` 和 `port`。 | ||
|
|
||
| 2. **Q: 运行时报错 “No module named ‘torch’”** | ||
| A: 请单独安装与你的Python及CUDA版本匹配的PyTorch。 | ||
|
|
||
| 3. **Q: 车辆控制不平稳,抖动严重** | ||
| A: 尝试调整 `config/default.yaml` 中的PID控制参数,或降低 `target_speed`。 | ||
|
|
||
| 4. **Q: 如何用自己的数据训练模型?** | ||
| A: 运行 `python main.py --mode collect` 采集数据,然后使用 `scripts/train_model.py` 进行训练。 | ||
|
|
||
| 5. **Q: ROS节点无法启动** | ||
| A: 确保ROS环境已正确配置 (`source /opt/ros/<distro>/setup.bash`),且所有ROS依赖已安装。 | ||
|
|
||
| --- | ||
| *更详细的API文档请使用 `mkdocs serve` 在本地构建和查看。* | ||
| MuJoCo 简易轮式小车智能代理 | ||
| 项目简介 | ||
| 一个基于PyTorch神经网络与MuJoCo物理引擎的简易轮式小车智能代理,实现了在连续控制环境下的感知、决策与控制一体化解决方案。项目采用模块化设计,便于在仿真环境中进行机器人控制算法研究、强化学习训练和控制系统验证。 | ||
|
|
||
| 功能特性 | ||
| 物理仿真:基于MuJoCo的高保真物理模拟,支持关节约束、碰撞检测和接触力计算 | ||
|
|
||
| 神经网络控制:使用PyTorch实现的深度强化学习或模仿学习策略 | ||
|
|
||
| 多模态感知:集成视觉、本体感知和惯性测量单元(IMU)等多种传感器输入 | ||
|
|
||
| 自适应控制:针对不同地形和任务需求调整控制策略 | ||
|
|
||
| 训练框架:提供完整的强化学习训练循环和评估接口 | ||
|
|
||
| 可视化工具:实时显示机器人状态、传感器数据和决策过程 | ||
|
|
||
| 环境要求 | ||
| 操作系统:Ubuntu 20.04/22.04(推荐)或 Windows 10/11(需配置WSL2) | ||
|
|
||
| Python:3.7 - 3.10(MuJoCo官方支持范围) | ||
|
|
||
| MuJoCo版本:2.3.0+(需要获取许可证) | ||
|
|
||
| 核心框架:PyTorch ≥ 1.9.0,MuJoCo-Py ≥ 2.3.0 | ||
|
|
||
| 渲染支持:OpenGL或EGL(GPU加速推荐) | ||
|
|
||
| 推荐硬件:支持CUDA的NVIDIA显卡(用于神经网络加速) | ||
|
|
||
| 快速开始 | ||
| 安装MuJoCo: | ||
|
|
||
| bash | ||
| # 获取MuJoCo许可证并下载 | ||
| pip install mujoco | ||
| # 设置许可证密钥(按官方指引操作) | ||
| 安装依赖: | ||
|
|
||
| bash | ||
| cd src/mujoco_wheeled_robot | ||
| pip install -r requirements.txt | ||
| 测试环境: | ||
|
|
||
| bash | ||
| python -c "import mujoco; print('MuJoCo version:', mujoco.__version__)" | ||
| 运行示例: | ||
|
|
||
| bash | ||
| # 启动随机策略演示 | ||
| python main.py --mode demo | ||
| # 开始训练 | ||
| python main.py --mode train --config config/training_config.yaml | ||
| 主要类和方法 | ||
| WheeledRobotEnv (环境类): | ||
|
|
||
| reset(): 重置环境到初始状态 | ||
|
|
||
| step(action): 执行动作并返回新状态、奖励和终止标志 | ||
|
|
||
| render(): 渲染当前环境状态 | ||
|
|
||
| PolicyNetwork (策略网络): | ||
|
|
||
| forward(observation): 根据观测计算动作分布 | ||
|
|
||
| sample(observation): 从分布中采样动作 | ||
|
|
||
| evaluate(observation, action): 计算动作的对数概率和熵 | ||
|
|
||
| DataLogger (数据记录器): | ||
|
|
||
| log_step(state, action, reward): 记录单步数据 | ||
|
|
||
| save_episode(episode_id): 保存回合数据 | ||
|
|
||
| export_trajectory(): 导出轨迹供分析使用 | ||
|
|
||
| 配置参数 | ||
| 关键配置位于 config/ 目录下: | ||
|
|
||
| yaml | ||
| # robot_config.yaml | ||
| robot: | ||
| mass: 2.5 # 小车质量 (kg) | ||
| wheel_radius: 0.1 # 轮子半径 (m) | ||
| max_velocity: 3.0 # 最大速度 (m/s) | ||
| control_frequency: 50 # 控制频率 (Hz) | ||
|
|
||
| # training_config.yaml | ||
| training: | ||
| algorithm: "sac" # 使用的算法 | ||
| total_timesteps: 1000000 | ||
| learning_rate: 3e-4 | ||
| gamma: 0.99 # 折扣因子 | ||
| batch_size: 256 | ||
| buffer_size: 1000000 | ||
| 数据输出 | ||
| 训练日志:TensorBoard格式,包含奖励曲线、策略熵等指标 | ||
|
|
||
| 模型检查点:定期保存的策略网络参数 | ||
|
|
||
| 轨迹数据:.npz格式,包含状态、动作、奖励序列 | ||
|
|
||
| 评估结果:性能指标统计(平均奖励、成功率等) | ||
|
|
||
| 视频记录:.mp4格式的演示视频 | ||
|
|
||
| 扩展建议 | ||
| 添加传感器:在模型文件中增加激光雷达或深度相机 | ||
|
|
||
| 更换任务:修改奖励函数实现不同任务(如避障、路径跟踪) | ||
|
|
||
| 集成真实数据:使用真实机器人数据微调仿真模型 | ||
|
|
||
| 多机器人协同:扩展环境支持多机器人交互 | ||
|
|
||
| 分布式训练:使用Ray等框架实现并行训练加速 | ||
|
|
||
| 模型预测控制:结合学习的动力学模型实现MPC控制器 | ||
|
|
||
| 常见问题 | ||
| Q: MuJoCo许可证如何获取? | ||
| A: 访问MuJoCo官网获取个人或机构许可证,按照官方指引配置环境变量。 | ||
|
|
||
| Q: 仿真运行速度慢 | ||
| A: 检查是否启用了GPU渲染,降低渲染分辨率,或使用 headless 模式进行训练。 | ||
|
|
||
| Q: 训练不稳定,奖励不收敛 | ||
| A: 尝试调整超参数(如学习率、折扣因子),增加网络容量,或检查奖励函数设计。 | ||
|
|
||
| Q: 如何导入自定义机器人模型? | ||
| A: 将 .xml 模型文件放入 models/ 目录,在环境初始化时指定模型路径。 | ||
|
|
||
| Q: 在Windows上运行问题 | ||
| A: 建议使用WSL2或Linux环境,Windows原生支持有限且可能遇到兼容性问题。 | ||
|
|
||
| Q: 模型无法学习有效策略 | ||
| A: 确保观测空间包含足够信息,检查动作空间是否合理,尝试更简单的任务作为起点。 |
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
需要删除