动态环境无人机导航的安全分层强化学习框架

商益铭,杜昌平,杨睿,方天睿,杜泽安,郑耀

Safe hierarchical reinforcement learning framework for dynamic UAV navigation

Yiming SHANG,Changping DU,Rui YANG,Tianrui FANG,Ze’an DU,Yao ZHENG

表 6 消融实验中的成功率、路径效率、平均转角和平均决策时间的对比

Tab.6 Comparison of success rate, path efficiency, average turning angle, and average decision time in ablation study

序号 +DWA +CBF +PDO Sc/% E $ {\theta }_{\text{avg}} $/(°) $ {T}_{\text{avg}} $/ms 0 — — — 64.3 0.739 21.41 1.87 1 √ — — 78.8 0.779 9.99 2.41 2 — √ — 85.6 0.893 17.72 2.03 3 — — √ 75.2 0.752 13.75 1.88 4 √ √ — 90.3 0.922 12.70 2.54 5 √ — √ 83.4 0.807 8.93 2.43 6 — √ √ 89.9 0.908 14.82 2.11 7 √ √ √ 95.7 0.962 9.41 2.57