281
个编辑
更改
跳到导航
跳到搜索
第622行：
第622行： + + + + + + + + +
→DQN 的应用领域
*输出动作：离散动作空间（如前进、左转、右转），适用于导航和避障任务。
*输出动作：离散动作空间（如前进、左转、右转），适用于导航和避障任务。
[[文件:应用于机器人路径规划.jpg|缩略图|基于DQN的机器人控制与路径规划算法示意图]]
[[文件:应用于机器人路径规划.jpg|缩略图|基于DQN的机器人控制与路径规划算法示意图]]
. 典型应用案例
室内导航与避障：
Tai等人（2017）<ref>{{cite journal|last1=Tai|first1=Lei|last2=Paolo|first2=Giuseppe|last3=Liu|first3=Ming|title=Virtual-to-real deep reinforcement learning: Continuous control of mobile robots for mapless navigation|journal=IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)|year=2017|pages=31-36|doi=10.1109/IROS.2017.8202134|publisher=IEEE}}</ref>：使用深度图像作为输入，训练轮式机器人在走廊环境中实现无地图导航，输出连续转向指令。
Zhang等人（2020）<ref>{{cite journal| last1 = Wang| first1 = Y.| last2 = Sun| first2 = J.| last3 = He| first3 = H.| last4 = Sun| first4 = C.| title = Deterministic policy gradient with integral compensator for robust quadrotor control| journal = IEEE Transactions on Systems, Man, and Cybernetics: Systems| volume = 50| issue = 10| pages = 3713–3725| year = 2020| doi = 10.1109/TSMC.2020.3002218| publisher = IEEE}}</ref>：通过特征迁移策略，将仿真环境训练的DQN模型迁移至真实机器人，解决仿真到现实（Sim-to-Real）的泛化问题。
Barron等人（2019）<ref> {{cite conference| last1 = Barron| first1 = T.| last2 = Whitehead| first2 = M.| last3 = Yeung| first3 = A.| title = Deep reinforcement learning in a 3-D blockworld environment| booktitle = Proceedings of the International Joint Conference on Artificial Intelligence (IJCAI)| pages = 1–6| year = 2016| publisher = IJCAI/AAAI Press}}</ref>：在3D虚拟环境中，采用RGB图像输入训练机器人，在复杂自然场景中实现优于深度输入的导航性能。
未知环境探索：
Haora等人（2020）<ref> {{cite journal| last1 = Li| first1 = H.| last2 = Zhang| first2 = Q.| last3 = Zhao| first3 = D.| title = Deep reinforcement learning-based automatic exploration for navigation in unknown environment| journal = IEEE Transactions on Neural Networks and Learning Systems| volume = 31| issue = 6| pages = 2064–2076| year = 2020| doi = 10.1109/TNNLS.2019.2930059| publisher = IEEE}}</ref>：提出自适应全卷积Q网络（AFCQN），结合局部地图信息，显著降低未知环境中的导航失败率。
无人机控制：
Kang等人（2019）<ref>{{cite journal|last1=Kang|first1=Kyunghoon|last2=Belkhale|first2=Suneel|last3=Kahn|first3=Gregory|last4=Abbeel|first4=Pieter|last5=Levine|first5=Sergey|title=Generalization through simulation: Integrating simulated and real data into deep reinforcement learning for vision-based autonomous flight|journal=IEEE International Conference on Robotics and Automation (ICRA)|year=2019|pages=6008-6014|doi=10.1109/ICRA.2019.8793701|publisher=IEEE}}</ref>：仅用单目视觉输入，训练四旋翼无人机在真实环境中实现动态避障。
===推荐系统中的策略优化===
===推荐系统中的策略优化===