|
|
| 多源威胁下高超声速飞行器分层在线决策规划 |
| Hierarchical Online Decision-Making and Planning for Hypersonic Vehicles under Multi-Source Threats |
| 投稿时间：2026-09-03 修订日期：2026-10-20 |
| DOI： |
| 中文关键词: 高超声速飞行器 多源威胁规避 分层决策 轨迹规划 近端策略优化 |
| English Keywords:hypersonic vehicles multi-source threats avoidance hierarchical decision-making trajectory planning proximal policy optimization |
| 基金项目:国家自然科学基金资助项目 |
|
| 摘要点击次数: 0 |
| 全文下载次数: 0 |
| 中文摘要: |
| 针对高超声速飞行器再入过程中禁飞区与拦截器并存下的在线决策与轨迹规划问题，提出一种双时间尺度分层决策规划方法。顶层将任务决策建模为部分可观测马尔可夫决策过程，基于近端策略优化（PPO）在线选择目标趋近、禁飞区绕飞与拦截器规避三类任务模式；底层调用对应元任务方法以更高频率生成连续制导指令，并经统一约束处理保证指令的可执行性。该架构实现任务决策与轨迹生成的层次化解耦，并通过状态反馈保留二者之间的动态耦合关系。蒙特卡罗仿真表明，所提方法成功率达89%，较规则基线提升21个百分点，平均单步决策耗时0.24ms，具有良好的约束满足能力、实时性与鲁棒性。 |
| English Summary: |
| A dual-time-scale hierarchical method is proposed for online decision-making and trajectory planning of hypersonic vehicle reentry amid coexisting no-fly zones and interceptors. The upper level formulates task decision-making as a Partially Observable Markov Decision Process and selects among target approach, no-fly zone bypass, and interceptor avoidance via a proximal policy optimization (PPO) policy. The lower level invokes the corresponding meta-task method at a higher rate to generate guidance commands, with unified constraint handling ensuring executability. The architecture decouples task decisions from trajectory generation while retaining their coupling through state feedback. Monte Carlo simulations show a success rate of 89%, 21 percentage points above a rule-based baseline, with a 0.24ms decision time, demonstrating favorable constraint satisfaction, real-time performance and robustness. |
| 查看/发表评论 下载PDF阅读器 |
