Pipeline.md

                  +-----------------------+
                  |    State Observation  |
                  | (车辆位置、任务需求等状态) |  ← [已正确包含状态信息]
                  +-----------------------+
                            |
                            v
                  +-----------------+
                  |   Actor Network |
                  | (策略网络 πθ(s)) |  ← [正确表示策略网络]
                  +-----------------+
                            |
        +--------------------+|+--------------------+
        | 概率: 本地计算 (p1) || 概率: 卸载到无人机 (p2) | ← [需添加地面服务器选项]
        +--------------------+|+--------------------+
                            |
                            v
            +--------------------------+
            |      动作选择 (Action)    |
            | p1/p2 卸载决策 + 资源分配  | ← [需添加资源分配部分]
            +--------------------------+
                            |
                            v
            +--------------------------+
            | 与环境交互 (Environment)  |
            |  - 计算时延 (Delay)      |
            |  - 资源分配 (Resources)   |  ← [正确的环境交互]
            |  - 奖励反馈 (Reward)     |
            +--------------------------+
                            |
                            v
                  +-----------------+
                  |   Critic Network |
                  | (价值网络 V(s))   |  ← [正确表示价值网络]
                  +-----------------+
                            |
                            v
            +-----------------------------+
            | PPO 更新策略:               |
            |  1. 策略目标: LCLIP(θ)      |  ← [正确的PPO更新机制]
            |  2. 限制更新幅度 (Clip ε)    |
            |  3. 使用优势函数 A(s,a)     |
            +-----------------------------+