class PrePretensioner:
    """
    Autoliv预收紧器系统
    
    工作流程：
    1. 安全ECU持续分析车辆动态
    2. 检测到潜在碰撞风险
    3. 命令预收紧器收紧安全带
    4. 消除松弛→乘员靠近最佳约束位置
    5. 减少碰撞前偏移（pre-crash excursion）
    
    关键参数：
    - 响应时间：<100ms（从检测到收紧）
    - 收紧力：200-400N
    - 收紧量：30-80mm
    """
    
    def __init__(self):
        self.trigger_conditions = {
            "brake_urgency": "紧急制动(减速度>7m/s²)",
            "ADAS_alert": "前向碰撞预警(FCW)",
            "vehicle_dynamics": "横摆率/侧向加速度异常",
            "AEB": "自动紧急制动激活"
        }
        
        self.performance = {
            "response_time_ms": 80,
            "tightening_force_N": 300,
            "tightening_distance_mm": 50,
            "excursion_reduction_mm": 75,
            "power_consumption_W": 15  # 仅触发时
        }
    
    def evaluate_trigger(self, vehicle_state: dict) -> dict:
        """评估是否需要触发预收紧"""
        triggers = []
        
        if vehicle_state.get("deceleration", 0) > 7:
            triggers.append("emergency_braking")
        
        if vehicle_state.get("fcw_active", False):
            triggers.append("forward_collision_warning")
        
        if vehicle_state.get("aeb_active", False):
            triggers.append("automatic_emergency_braking")
        
        if abs(vehicle_state.get("yaw_rate", 0)) > 10:
            triggers.append("lateral_instability")
        
        should_trigger = len(triggers) > 0
        
        return {
            "trigger": should_trigger,
            "reasons": triggers,
            "confidence": min(1.0, len(triggers) * 0.4)
        }


pt = PrePretensioner()

# 测试场景
scenarios = [
    {"name": "正常巡航", "state": {"deceleration": 0.5, "fcw_active": False, "aeb_active": False, "yaw_rate": 0.5}},
    {"name": "紧急制动", "state": {"deceleration": 8.5, "fcw_active": True, "aeb_active": True, "yaw_rate": 1.0}},
    {"name": "急转弯", "state": {"deceleration": 1.0, "fcw_active": False, "aeb_active": False, "yaw_rate": 12.0}},
]

for s in scenarios:
    result = pt.evaluate_trigger(s["state"])
    print(f"{s['name']:<12} 触发: {'✅' if result['trigger'] else '❌'}  原因: {result['reasons']}")

import torch
import torch.nn as nn

class SeatbeltMisuseDetector(nn.Module):
    """
    安全带误用检测器
    
    输入：
    1. 摄像头图像特征（安全带可见性+位置）
    2. 张力传感器读数
    3. 织物压力分布
    4. 扣具状态
    
    输出：
    1. 佩戴状态（正确/手臂下/背后/未扣/过松）
    2. 置信度
    3. 建议动作
    """
    
    def __init__(self):
        super().__init__()
        
        # 摄像头特征（简化）
        self.camera_encoder = nn.Sequential(
            nn.Linear(64, 128), nn.GELU(), nn.Linear(128, 64)
        )
        
        # 张力传感器
        self.tension_encoder = nn.Sequential(
            nn.Linear(5, 32), nn.GELU(), nn.Linear(32, 16)
        )
        
        # 压力分布
        self.pressure_encoder = nn.Sequential(
            nn.Linear(20, 64), nn.GELU(), nn.Linear(64, 16)
        )
        
        # 融合分类器
        self.classifier = nn.Sequential(
            nn.Linear(64 + 16 + 16 + 1, 64),  # +1 for buckle status
            nn.GELU(),
            nn.Linear(64, 6)  # 6类: correct/arm/behind/unbuckled/loose/child
        )
    
    def forward(self, camera_feat, tension, pressure, buckle):
        c = self.camera_encoder(camera_feat)
        t = self.tension_encoder(tension)
        p = self.pressure_encoder(pressure)
        b = buckle  # (B, 1) 扣具状态
        
        x = torch.cat([c, t, p, b], dim=-1)
        logits = self.classifier(x)
        
        return {
            "classification": logits,
            "misuse_detected": logits.argmax(dim=-1) != 0,
            "confidence": torch.softmax(logits, dim=-1).max(dim=-1).values
        }


# 测试
detector = SeatbeltMisuseDetector()

scenarios = {
    "正确佩戴": {"tension": torch.tensor([[30, 28, 31, 29, 30]]), "buckle": torch.tensor([[1.0]]), "expected": "correct"},
    "手臂下方": {"tension": torch.tensor([[15, 14, 16, 15, 14]]), "buckle": torch.tensor([[1.0]]), "expected": "arm"},
    "未扣": {"tension": torch.tensor([[0, 0, 0, 0, 0]]), "buckle": torch.tensor([[0.0]]), "expected": "unbuckled"},
    "过松": {"tension": torch.tensor([[5, 5, 6, 5, 5]]), "buckle": torch.tensor([[1.0]]), "expected": "loose"},
}

for name, data in scenarios.items():
    camera_feat = torch.randn(1, 64)  # 简化
    pressure = torch.randn(1, 20)
    result = detector(camera_feat, data["tension"], pressure, data["buckle"])
    pred = ["correct", "arm", "behind", "unbuckled", "loose", "child"][result["classification"].argmax(1).item()]
    print(f"{name:<12} 预测: {pred:<12} 误用: {'❌' if result['misuse_detected'].item() else '✅'}")