"""
Tobii 提到的 Interior Sensing Corner Cases 分类

核心挑战：室内感知必须覆盖巨大的多样性
- 不同的人（年龄、性别、种族、体型）
- 不同的行为（驾驶习惯、坐姿、互动方式）
- 不同的物理特征（眼镜、口罩、帽子、胡子）
- 不同的光照条件（白天、夜晚、隧道、逆光）
- 不同的摄像头位置（方向盘后、仪表盘、顶控台）
- 不同的舱内布局（轿车、SUV、MPV、卡车）
- 不同的遮挡场景（手遮脸、安全带遮胸、太阳镜遮眼）
- 罕见场景（突发疾病、异常姿态、儿童遗留）
"""

from dataclasses import dataclass
from typing import List
from enum import Enum


class CornerCaseCategory(Enum):
    """Corner Case 分类"""
    PERSON_DIVERSITY = "人员多样性"
    BEHAVIOR = "行为差异"
    APPEARANCE = "外观遮挡"
    LIGHTING = "光照条件"
    CAMERA_POSITION = "摄像头位置"
    CABIN_LAYOUT = "舱内布局"
    OCCLUSION = "遮挡场景"
    RARE_EVENT = "罕见事件"


@dataclass
class CornerCase:
    """Corner Case 定义"""
    category: CornerCaseCategory
    description: str
    difficulty: int  # 1-5, 5最难
    data_coverage: float  # 0-1, 当前数据覆盖率
    ims_priority: str  # P0/P1/P2


# Tobii 提到的核心 corner cases
TOBII_CORNER_CASES = [
    # 人员多样性
    CornerCase(
        category=CornerCaseCategory.PERSON_DIVERSITY,
        description="不同种族/肤色的面部关键点检测",
        difficulty=3, data_coverage=0.7,
        ims_priority="P0"
    ),
    CornerCase(
        category=CornerCaseCategory.PERSON_DIVERSITY,
        description="儿童（3-12岁）面部特征差异",
        difficulty=4, data_coverage=0.4,
        ims_priority="P0"
    ),
    CornerCase(
        category=CornerCaseCategory.PERSON_DIVERSITY,
        description="老年人面部皱纹导致关键点偏移",
        difficulty=3, data_coverage=0.5,
        ims_priority="P1"
    ),
    
    # 外观遮挡
    CornerCase(
        category=CornerCaseCategory.APPEARANCE,
        description="墨镜遮挡眼睛（PERCLOS 失效）",
        difficulty=5, data_coverage=0.3,
        ims_priority="P0"
    ),
    CornerCase(
        category=CornerCaseCategory.APPEARANCE,
        description="口罩遮挡下半脸",
        difficulty=4, data_coverage=0.6,
        ims_priority="P0"
    ),
    CornerCase(
        category=CornerCaseCategory.APPEARANCE,
        description="安全带遮挡胸部/肩部",
        difficulty=3, data_coverage=0.5,
        ims_priority="P1"
    ),
    
    # 光照条件
    CornerCase(
        category=CornerCaseCategory.LIGHTING,
        description="逆光导致面部全黑",
        difficulty=5, data_coverage=0.4,
        ims_priority="P0"
    ),
    CornerCase(
        category=CornerCaseCategory.LIGHTING,
        description="隧道进出光照突变",
        difficulty=4, data_coverage=0.3,
        ims_priority="P1"
    ),
    CornerCase(
        category=CornerCaseCategory.LIGHTING,
        description="屏下摄像头低透过率",
        difficulty=5, data_coverage=0.2,
        ims_priority="P1"
    ),
    
    # 罕见事件
    CornerCase(
        category=CornerCaseCategory.RARE_EVENT,
        description="驾驶员突发疾病（癫痫/心梗）",
        difficulty=5, data_coverage=0.1,
        ims_priority="P1"
    ),
    CornerCase(
        category=CornerCaseCategory.RARE_EVENT,
        description="儿童独自在后座（CPD）",
        difficulty=4, data_coverage=0.5,
        ims_priority="P0"
    ),
    CornerCase(
        category=CornerCaseCategory.RARE_EVENT,
        description="异常姿态（OOP - 脚在仪表盘上）",
        difficulty=4, data_coverage=0.3,
        ims_priority="P0"
    ),
]


def assess_data_gaps(cases: List[CornerCase]):
    """评估数据覆盖缺口"""
    print("=== Corner Case 数据覆盖评估 ===\n")
    for case in cases:
        gap = 1 - case.data_coverage
        status = "✅ 充分" if gap < 0.3 else \
                 "⚠️ 不足" if gap < 0.6 else \
                 "❌ 严重不足"
        print(f"  [{case.ims_priority}] {case.description}")
        print(f"    难度: {'★'*case.difficulty} "
              f"覆盖: {case.data_coverage*100:.0f}% "
              f"缺口: {gap*100:.0f}% {status}\n")


assess_data_gaps(TOBII_CORNER_CASES)

"""
L2 → L3 座舱感知需求升级

L2: 驾驶员必须随时准备接管
L3: 系统可以完全控制，驾驶员可在特定条件下脱离驾驶

关键差异：
- L2: 检测驾驶员是否在监控道路
- L3: 评估驾驶员是否具备接管能力
- L3: 需要更复杂的就绪度评估
"""

L2_VS_L3 = {
    '监控目标': {
        'L2': '驾驶员是否在看路',
        'L3': '驾驶员是否具备接管能力'
    },
    '检测项': {
        'L2': ['疲劳', '分心', '视线方向'],
        'L3': ['疲劳', '分心', '视线方向', 
               '功能就绪度', '接管准备度',
               '认知负荷', '情绪状态']
    },
    '摄像头需求': {
        'L2': '1个（单摄像头 DMS+OMS）',
        'L3': '2+个（多摄像头 + 传感器融合）'
    },
    '决策复杂度': {
        'L2': '二元警告（警告/不警告）',
        'L3': '分级干预（警告→辅助→接管）'
    },
    'ADAS协同': {
        'L2': '弱耦合（DMS 触发警告）',
        'L3': '强耦合（DMS 影响 ADAS 控制策略）'
    },
    '数据需求': {
        'L2': '标准场景覆盖',
        'L3': '极端 corner case + 长尾覆盖'
    }
}

for dimension, values in L2_VS_L3.items():
    print(f"\n=== {dimension} ===")
    print(f"  L2: {values['L2']}")
    print(f"  L3: {values['L3']}")

"""
全景感知：行为意图解码

核心思想：
- 舱内感知：检测"驾驶员在看哪里"
- ADAS 外部感知：检测"外部有什么"
- 融合解码：理解"为什么驾驶员在看那里"

示例：
- 驾驶员看右后视镜 + 右车道有车 → 准备变道（正常）
- 驾驶员看手机 + 前方有障碍 → 分心（危险）
- 驾驶员闭眼 + 直路无车 → 疲劳（警告）
- 驾驶员闭眼 + 前方有行人 → 紧急（立即干预）
"""

from dataclasses import dataclass
from typing import Optional
from enum import IntEnum


class GazeTarget(IntEnum):
    ROAD = 0         # 看路
    MIRROR = 1       # 看后视镜
    PHONE = 2        # 看手机
    INFOTAINMENT = 3 # 看中控屏
    CLOSED = 4       # 闭眼
    PASSENGER = 5    # 看乘客


class RoadContext(IntEnum):
    CLEAR = 0         # 前方无障碍
    VEHICLE_AHEAD = 1 # 前方有车
    PEDESTRIAN = 2    # 前方有行人
    CURVE = 3         # 弯道
    INTERSECTION = 4  # 路口


@dataclass
class HolisticAssessment:
    """全景感知评估"""
    gaze: GazeTarget
    road: RoadContext
    intent: str
    risk_level: int  # 0-3
    action: str


def assess_intent(gaze: GazeTarget, 
                  road: RoadContext) -> HolisticAssessment:
    """
    融合舱内感知和 ADAS 外部感知
    解码驾驶员行为意图
    """
    # 意图矩阵
    INTENT_MATRIX = {
        (GazeTarget.ROAD, RoadContext.CLEAR): 
            ("正常驾驶", 0, "无动作"),
        (GazeTarget.ROAD, RoadContext.VEHICLE_AHEAD): 
            ("跟随前车", 0, "无动作"),
        (GazeTarget.ROAD, RoadContext.PEDESTRIAN): 
            ("注意到行人", 0, "准备制动"),
        (GazeTarget.MIRROR, RoadContext.CLEAR): 
            ("检查后方", 0, "可能变道"),
        (GazeTarget.MIRROR, RoadContext.VEHICLE_AHEAD): 
            ("检查后方+前车", 1, "维持车距"),
        (GazeTarget.PHONE, RoadContext.CLEAR): 
            ("分心（低风险）", 1, "一级警告"),
        (GazeTarget.PHONE, RoadContext.VEHICLE_AHEAD): 
            ("分心（中风险）", 2, "二级警告"),
        (GazeTarget.PHONE, RoadContext.PEDESTRIAN): 
            ("分心（高风险）", 3, "紧急制动"),
        (GazeTarget.CLOSED, RoadContext.CLEAR): 
            ("疲劳（低风险）", 1, "疲劳警告"),
        (GazeTarget.CLOSED, RoadContext.VEHICLE_AHEAD): 
            ("疲劳（中风险）", 2, "疲劳警告+减速"),
        (GazeTarget.CLOSED, RoadContext.PEDESTRIAN): 
            ("疲劳（高风险）", 3, "紧急制动+接管"),
        (GazeTarget.INFOTAINMENT, RoadContext.CLEAR): 
            ("操作中控（低风险）", 0, "无动作"),
        (GazeTarget.INFOTAINMENT, RoadContext.PEDESTRIAN): 
            ("操作中控（高风险）", 2, "二级警告"),
    }
    
    intent, risk, action = INTENT_MATRIX.get(
        (gaze, road), 
        ("未知", 1, "监控")
    )
    
    return HolisticAssessment(
        gaze=gaze, road=road,
        intent=intent, risk_level=risk, action=action
    )


# 测试
if __name__ == "__main__":
    print("=== 全景感知意图解码 ===\n")
    
    scenarios = [
        (GazeTarget.MIRROR, RoadContext.CLEAR, 
         "驾驶看后视镜+直路"),
        (GazeTarget.PHONE, RoadContext.PEDESTRIAN, 
         "驾驶看手机+前方行人"),
        (GazeTarget.CLOSED, RoadContext.PEDESTRIAN, 
         "驾驶闭眼+前方行人"),
        (GazeTarget.ROAD, RoadContext.PEDESTRIAN, 
         "驾驶看路+前方行人"),
    ]
    
    for gaze, road, desc in scenarios:
        result = assess_intent(gaze, road)
        print(f"场景: {desc}")
        print(f"  意图: {result.intent}")
        print(f"  风险: L{result.risk_level}")
        print(f"  动作: {result.action}\n")