Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
44 changes: 44 additions & 0 deletions src/HCL_interaction_task/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -30,6 +30,9 @@
- **反应时间记录**:精确记录从面板目标出现到食指选择的反应时长,用于交互性能分析
- **选择有效性判定**:自动检测食指与目标区域的接触状态,判断选择有效性
- **动态目标生成**:支持随机/序列生成面板目标,模拟复杂选择任务
- **任务逻辑闭环**:通过task.py实现奖励机制、超时判断、任务重置的完整逻辑闭环
- **友好状态提示**:终端实时输出任务进度、奖励数值、碰撞状态,便于调试与分析



## 目录结构
Expand All @@ -48,6 +51,7 @@ HCL_interaction_task/
│ ├── simulation.xml # 面板功能模型定义
│ ├── main.py # 面板模块程序入口
│ ├── simulator.py # 面板仿真核心逻辑
│ ├── task.py # 面板任务逻辑核心
│ └── assets/ # 面板模块模型资源(含面板纹理)
├── README.md # 项目说明文档
└── requirements.txt # 项目依赖清单
Expand All @@ -65,6 +69,41 @@ HCL_interaction_task/
- 关键定义:面板碰撞属性、目标区域触发条件、手指接触检测参数



## 核心模块说明

### 面板任务逻辑核心(`mobl_arms_index_choice_reaction/task.py`)
#### 1. 模块作用
作为面板选择反应模块的「大脑」,负责串联仿真流程与交互逻辑,实现从任务初始化到结果判定的完整闭环,无需侵入修改仿真核心代码,保持模块解耦性。

#### 2. 核心功能
| 功能项 | 详细描述 |
|-----------------------|--------------------------------------------------------------------------|
| 任务初始化/重置 | 支持任务参数加载与状态重置,实现多轮任务自动循环运行 |
| 指尖-目标碰撞检测 | 精准识别食指远节(`hand_2distph`)与面板目标(`button-x`)的接触状态 |
| 奖励机制设计 | 正确选择目标奖励(默认10.0)、超时失败扣分(默认-5.0)、动作成本惩罚(可选) |
| 超时判断 | 基于配置的超时步数/时间,自动判定任务失败并触发重置 |
| 实时状态反馈 | 每50步输出任务进度、剩余步数、当前奖励,便于实时监控仿真过程 |
| 兼容无执行器模型 | 对无肌肉/执行器驱动的模型提供友好提示,不影响任务核心逻辑运行 |

#### 3. 关键类与方法
- **`ChoicePanelTask` 类**:任务逻辑封装核心,关联仿真器实例与配置参数
- `__init__`:初始化任务参数(目标按钮、奖励值、超时步数)
- `reset()`:重置任务状态,用于多轮任务循环
- `update()`:每步仿真调用,更新任务状态(碰撞检测、奖励计算、超时判断)
- `_check_button_contact()`:内部核心方法,实现指尖与目标的碰撞检测

#### 4. 配置依赖
任务参数完全读取自`config.yaml`,支持灵活配置,无需修改代码:
```yaml
# 任务相关配置(已整合到config.yaml)
target_button: 0 # 目标按钮ID(0-3对应红/绿/蓝/黄)
button_reward: 10.0 # 正确选择奖励值
effort_cost: 0.01 # 动作成本惩罚系数(有执行器时生效)
timeout: 600 # 任务超时步数(对应仿真时长)
```


## 系统要求

- Ubuntu 22.04 LTS (ROS 2 Humble)
Expand Down Expand Up @@ -131,6 +170,11 @@ panel_pos: [0.5, 0, 0.6] # 面板位置
target_count: 3 # 面板目标数量
target_size: 0.05 # 目标区域大小
reaction_timeout: 5.0 # 反应超时时间(秒)
# 任务逻辑配置(新增)
target_button: 0 # 目标按钮ID(0-3)
button_reward: 10.0 # 正确选择奖励值
effort_cost: 0.01 # 动作成本惩罚系数
timeout: 600 # 任务超时步数
```


Expand Down
Original file line number Diff line number Diff line change
@@ -1,2 +1,8 @@
# 原有配置
dt: 0.01 # 仿真步长(调慢一点,看得清)
render_mode: "human" # 开启可视化!
render_mode: "human" # 开启可视化!

target_button: 0 # 目标按钮(可选0/1/2/3,对应红/绿/蓝/黄)
button_reward: 10.0 # 选对按钮的奖励值
effort_cost: 0.01 # 动作成本扣分系数(模型有执行器时生效)
timeout: 600 # 任务超时步数(300步≈9秒,可根据需要调整)
59 changes: 57 additions & 2 deletions src/HCL_interaction_task/mobl_arms_index_choice_reaction/main.py
Original file line number Diff line number Diff line change
@@ -1,13 +1,68 @@
import mujoco
from simulator import IndexSimulator
from task import ChoicePanelTask # 新增:导入任务类
import yaml # 新增:加载配置文件

def main():
# 1. 配置路径
config_path = "config.yaml"
model_path = "simulation.xml"

# 2. 初始化仿真器+运行
# 新增:加载配置文件(给任务传参)
with open(config_path, 'r', encoding='utf-8') as f:
config = yaml.safe_load(f)

# 2. 初始化仿真器+任务实例
sim = IndexSimulator(config_path, model_path)
sim.run_simulation() # 所有打印都在run_simulation里,这里不用额外打印
task = ChoicePanelTask(config, sim) # 关联配置和仿真器

# 3. 重置仿真和任务
sim.reset()
task.reset()

# 4. 仿真循环(整合任务逻辑,替代sim.run_simulation())
try:
# 先启动可视化viewer(和原仿真器逻辑一致)
try:
sim.viewer = mujoco.viewer.launch_passive(sim.model, sim.data)
# 优化视角
sim.viewer.cam.azimuth = 135
sim.viewer.cam.elevation = -15
sim.viewer.cam.distance = 0.6
sim.viewer.cam.lookat = [0.45, -0.15, 0.8]
print("✅ 可视化窗口启动成功!能看到手臂慢速摆动")
except Exception as e:
print(f"⚠️ 可视化启动失败(不影响动作):{e}")
print("📢 无窗口模式:终端打印关节角度,确认动作真实发生!")

# 循环执行仿真和任务更新
while sim.is_running:
# 仿真步进(模型运动)
sim.step()
# 任务状态更新(判断成败、计算奖励)
task_status = task.update()

# 如果任务完成(成功/超时),重置仿真和任务
if task_status["done"]:
print("\n🔄 准备开始新一轮任务...")
sim.reset()
task.reset()

# 渲染同步(保持窗口流畅)
if sim.viewer:
sim.viewer.sync()
time.sleep(0.001)

except KeyboardInterrupt:
print("\n\n⚠️ 检测到Ctrl+C,正在优雅退出仿真...")
sim.is_running = False
finally:
# 关闭资源
sim.close()
print(f"\n✅ 仿真正常结束!共运行{sim.current_step}步")

# 新增:导入time(用于渲染延时)
import time

if __name__ == "__main__":
main()
Original file line number Diff line number Diff line change
Expand Up @@ -32,20 +32,20 @@ def __init__(self, config_path, model_path):
15: {
"name": "shoulder_rot (肩关节旋转)",
"range": (-1.57, 0.349),
"step": 0.02,
"current_step_dir": 0.02 # 运动方向
"step": 0.05,
"current_step_dir": 0.05 # 运动方向
},
16: {
"name": "elbow_flexion (肘关节弯曲)",
"range": (0, 2.26893),
"step": 0.03,
"current_step_dir": 0.03
"step": 0.08,
"current_step_dir": 0.08
},
17: {
"name": "pro_sup (前臂旋前旋后)",
"range": (-1.5708, 1.5708),
"step": 0.015,
"current_step_dir": 0.015
"step": 0.04,
"current_step_dir": 0.04
}
}

Expand Down
82 changes: 82 additions & 0 deletions src/HCL_interaction_task/mobl_arms_index_choice_reaction/task.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,82 @@
import numpy as np

class ChoicePanelTask:
def __init__(self, config, sim):
self.config = config
self.sim = sim # 关联你的仿真器实例
# 从配置中读取任务参数(后续在config.yaml中配置)
self.target_button = config.get("target_button", 0) # 默认目标按钮0
self.button_reward = config.get("button_reward", 10.0) # 选对奖励
self.effort_cost = config.get("effort_cost", 0.01) # 动作成本系数
self.timeout_steps = config.get("timeout", 300) # 超时步数

self.step_count = 0 # 记录当前任务步数
self.done = False # 任务是否完成
self.reward = 0.0 # 当前奖励值

def reset(self):
"""重置任务状态(每次仿真/任务重新开始时调用)"""
self.step_count = 0
self.done = False
self.reward = 0.0
print(f"\n🔄 任务已重置!目标按钮:button-{self.target_button},超时步数:{self.timeout_steps}")
return {"reward": self.reward, "done": self.done}

def update(self):
"""每一步仿真都调用,更新任务状态(核心逻辑)"""
if self.done: # 任务已完成,直接返回状态
return {"reward": self.reward, "done": self.done}

self.step_count += 1
self.reward = 0.0

# 1. 检测:指尖是否碰到目标按钮
is_touch_target = self._check_button_contact(self.target_button)
if is_touch_target:
self.reward += self.button_reward # 选对按钮,添加奖励
self.done = True
print(f"\n🎉 任务成功!指尖碰到button-{self.target_button},奖励:{self.reward:.2f}")

# 2. 动作成本扣分(兼容执行器为0的情况)
ctrl_effort = np.sum(np.abs(self.sim.data.ctrl)) if self.sim.model.nu > 0 else 0.0
effort_penalty = self.effort_cost * ctrl_effort
self.reward -= effort_penalty
if self.sim.model.nu == 0 and self.step_count % 50 == 0:
print(f"\n💡 提示:当前模型无执行器,动作成本扣分暂不生效(后续添加执行器后自动启用)")

# 3. 超时判断:超过步数未完成则任务失败
if self.step_count >= self.timeout_steps:
self.done = True
self.reward -= 5.0 # 超时额外扣分
print(f"\n⏰ 任务超时失败!未在{self.timeout_steps}步内碰到button-{self.target_button},最终奖励:{self.reward:.2f}")

# 4. 定期打印任务状态(每50步)
if self.step_count % 50 == 0 and not self.done:
print(f"\n💰 step{self.step_count} 任务状态:")
print(f" 是否碰到目标按钮:{'是' if is_touch_target else '否'}")
print(f" 当前奖励:{self.reward:.2f}")
print(f" 剩余步数:{self.timeout_steps - self.step_count}")

return {"reward": self.reward, "done": self.done}

def _check_button_contact(self, button_id):
"""检测指尖(hand_2distph)是否碰到指定按钮(button-x)"""
finger_geom_name = self.sim.finger_geom_name # 从仿真器获取指尖geom名称
target_button_name = f"button-{button_id}" # 目标按钮geom名称

# 遍历所有碰撞对,判断是否包含「指尖」和「目标按钮」
for i in range(self.sim.data.ncon):
contact = self.sim.data.contact[i]
# 获取两个碰撞geom的名称
geom1_id = contact.geom1
geom2_id = contact.geom2
geom1_name = mujoco.mj_id2name(self.sim.model, mujoco.mjtObj.mjOBJ_GEOM, geom1_id)
geom2_name = mujoco.mj_id2name(self.sim.model, mujoco.mjtObj.mjOBJ_GEOM, geom2_id)

# 判断是否是「指尖-目标按钮」的碰撞
if (finger_geom_name in [geom1_name, geom2_name]) and (target_button_name in [geom1_name, geom2_name]):
return True
return False

# 导入mujoco(避免内部函数报错)
import mujoco
Loading