Path planning is a core issue in fields such as robotics, autonomous driving, and intelligent transportation. Aiming at the path planning problem in complex dynamic environments, the combination of RL (reinforcement learning) and TL (transfer learning) provide new ideas. The limitations of traditional path planning methods were first reviewed, followed by a systematic analysis of the key technologies of RL and TL in path planning as well as their integration potential. It is found that the dynamic decision-making advantage of RL and the cross-domain knowledge reuse capability of TL form a complement. Their integrated architecture can effectively resolve the contradiction between sample efficiency and generalization of path planning in complex dynamic environments. Meanwhile, domain similarity quantification, negative transfer suppression, and safety interpretability were identified as the core research pain points in this field. Finally, future research directions were prospected from aspects such as efficient knowledge transfer and multi-task collaborative optimization.
LI Meng-meng, ZHANG Min, SHANG Zhi-gang, LI Zhi-hui, YANG Li-fang.
Survey on Path Planning Research Based on Reinforcement Learning and Transfer Learning[J].
Science Technology and Engineering, 2026, 26(26): 11107-11116 DOI:10.12404/j.issn.1671-1815.2504462
在智能体自主导航体系中,路径规划是支撑机器人自主运动与任务执行的关键技术,其目标是在满足碰撞约束和运动学约束的条件下,为智能体寻找从起点到目标点的最优或可行路径[1-3]。传统路径规划方法中,A*算法[4]和快速扩展随机树(rapidly-exploring random tree,RRT)算法[5]在已知静态环境或高维构型空间中具有代表性。然而,随着实际应用场景向复杂化、动态化发展,这类方法在动态障碍、高维状态空间以及部分可观测条件下的局限性逐步凸显。
LaValle[5]于1998年提出快速扩展随机树(rapidly-exploring random tree,RRT)算法。该算法通过随机采样和树结构增量扩展探索构型空间,能够在高维、非结构化环境中较快获得可行路径。RRT并不保证所得路径最优,其路径质量与搜索效率受采样策略、步长及树扩展规则影响[18];在不同任务之间迁移时,这些参数通常需要重新调整,限制了其跨场景适用性[19]。
在路径规划任务中,离散动作空间通常将智能体的行为划分为有限个可枚举的基础动作,对动作决策的实时性与策略收敛效率有较高要求。针对离散动作空间的特性,强化学习算法通过构建离散状态-动作对的价值评估体系,实现对路径决策策略的优化。下面介绍典型的Q-learning和双深度Q网络(double deep Q-network, DDQN)两种算法。
ZhangRongxia, WuChangxu, SunTongchao, et al. Research progress of deep reinforcement learning and its application in path planning[J]. Computer Engineering and Applications, 2021, 57(19): 44-56.
GuoYuli, YangWenfan, ZhangHuan, et al. Research status of path planning algorithms for guide robots[J]. Chinese Medical Equipment Journal, 2025, 46(2): 92-101.
[7]
HartP E, NilssonN J, RaphaelB. A formal basis for the heuristic determination of minimum cost paths[J]. IEEE Transactions on Systems Science and Cybernetics, 1968, 4(2): 100-107.
[8]
LaValleS M. Rapidly-exploring random trees: a new tool for path planning[M]. Ames: Iowa State University, 1998.
WuDingjie, LiuGerui, XieWeizheng. Rerouting path planning based on the improved ant colony algorithm with A*[J]. Science Technology and Engineering, 2025, 25(32): 14053-14061.
[20]
WangK, LiS, ZhangM, et al. Research on AGV path optimization based on an improved A* and DWA fusion algorithm[J]. Forests, 2026, 17(1): 31.
[21]
FoxD, BurgardW, ThrunS. The dynamic window approach to collision avoidance[J]. IEEE Robotics & Automation Magazine, 1997, 4(1): 23-33.
JiaQianxi, ZhaoXu, MengZhijun. Path planning algorithm in dynamic environment based on an improved dynamic window approach[J]. Science Technology and Engineering, 2024, 24(15): 6313-6319.
WangShuo, DuanRongkai, LiaoYuhe. Improved research on rapidly-exploring random tree algorithm in robot path planning[J]. Journal of Xi’an Jiaotong University, 2022, 56(7): 1-8.
PanGuoqian, ZhouXinzhi. Research on mobile robot path planning algorithm based on heuristic reinforcement learning[J]. Modern Computer, 2022, 28(10): 57-61.
ZhangHaoyu, XiongKai. Research on gait control of quadruped robots based on proximal policy optimization algorithm[J]. Aerospace Control and Application, 2019, 45(3): 53-58.
XiaWei, LiHuiyun. Autonomous driving strategy learning method based on deep reinforcement learning[J]. Journal of Integration Technology, 2017, 6(3): 29-34.
[38]
WatkinsC J C H. Learning from delayed rewards[D]. Cambridge: University of Cambridge, 1989.
MnihV, KavukcuogluK, SilverD, et al. Human-level control through deep reinforcement learning[J]. Nature, 2015, 518: 529-533.
[41]
van HasseltH, GuezA, SilverD. Deep reinforcement learning with double Q-learning[C]//Proceedings of the 29th AAAI Conference on Artificial Intelligence. Austin: AAAI Press, 2015: 2094-2100.
[42]
KouloughliI, FecianeM K, FandiW. A Q-learning approach for optimizing the retrieval time in an AS/RS[J]. The International Journal of Advanced Manufacturing Technology, 2025, 140: 5515-5534.
[43]
Antony VijeshV, ShreyasS R. A weighted smooth Q-Learning algorithm[J]. IEEE Control Systems Letters, 2025, 9: 21-26.
[44]
BarghashA, AbuznaidA. Advances in Q-learning: real-time optimization of multi-distant transportation systems[J]. Applied Sciences, 2025, 15(17): 9493.
[45]
ZhengH, ZhangJ L, JiangM Y, et al. Embodied escaping: end-to-end reinforcement learning for robot navigation in narrow environment[C]// IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). Hangzhou: IEEE, 2025: 9719-9726.
[46]
LillicrapT P, HuntJ J, PritzelA, et al. Continuous control with deep reinforcement learning[C]// 3rd International Conference on Learning Representations. San Diego: ICLR, 2016: arXiv: 1509.02971v5.
MnihV, BadiaA P, MirzaM, et al. Asynchronous methods for deep reinforcement learning[C]// Proceedings of the 33rd International Conference on Machine Learning. New York: PMLR, 2016: 1928-1937.
[49]
Garrido-CastañedaS I, VasquezJ I, Antonio-CruzM. Coverage path planning using actor-critic deep reinforcement learning[J]. Sensors, 2025, 25(5): 1592.
[50]
ZhouY, YangY, PanB D, et al. Deep reinforcement learning for real-time airport emergency evacuation using asynchronous advantage actor-critic (A3C) algorithm[J]. Mathematics, 2025, 13(14): 2269.
[51]
ZhuW, LiuX, LiuY, et al. RT-A3C: real-time asynchronous advantage actor-critic for optimally defending malicious attacks in edge-enabled industrial internet of things[J]. Journal of Information Security and Applications, 2025, 91: 104073.
[52]
DuanJ, LiS E, GuanY, et al. Distributional soft actor-critic with three refinements[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025, 47(5): 3935-3946.
[53]
GaoF, WuW, GaoW, et al. Flying on point clouds: online trajectory generation and autonomous navigation for quadrotors in cluttered environments[J]. Journal of Field Robotics, 2019, 36: 710-733.
[54]
NgA Y, RussellS J. Algorithms for inverse reinforcement learning[C]// Proceedings of the 17th International Conference on Machine Learning. San Francisco: Morgan Kaufmann, 2000: 663-670.
[55]
ZiebartB D, MaasA L, BagnellJ A, et al. Maximum entropy inverse reinforcement learning[C]// Proceedings of the 23rd AAAI Conference on Artificial Intelligence. Chicago: AAAI Press, 2008: 1433-1438.
[56]
HoJ, ErmonS. Generative adversarial imitation learning[C]// Advances in Neural Information Processing Systems. San Diego: NIPS, 2016: 4565-4573.
ZhangChao, BaiWensong, DuXin, et al. A survey of imitation learning: traditional and new advances[J]. Journal of Image and Graphics, 2023, 28(6): 1585-1607.
ZhangLihua, LiuQuan, HuangZhigang, et al. A survey of inverse reinforcement learning[J]. Journal of Software, 2023, 34(10): 4772-4803.
[63]
WulfmeierM, RaoD, WangD Z, et al. Large-scale cost function learning for path planning using deep inverse reinforcement learning[J]. The International Journal of Robotics Research, 2017, 36(10): 1073-1087.
[64]
SuttonR S, PrecupD, SinghS. Between MDPs and semi-MDPs: a framework for temporal abstraction in reinforcement learning[J]. Artificial Intelligence, 1999, 112(1/2): 181-211.
HuangZhigang, LiuQuan, ZhangLihua, et al. Research and development of deep hierarchical reinforcement learning[J]. Journal of Software, 2023, 34(2): 733-760.
[67]
YanK, LiuD, LiuZ, et al. An inspection path planning approach based on hierarchical reinforcement learning[J]. Journal of Manufacturing Systems, 2026, 85: 34-50.
YinChangsheng, YangRuopeng, ZhuWei, et al. A survey of multi-agent hierarchical reinforcement learning[J]. CAAI Tran-sactions on Intelligent Systems, 2020, 15(4): 646-655.
[70]
van EssenR, van HentenE, KootstraG. UAV-based path planning for efficient localization of non-uniformly distributed weeds using prior knowledge: a inforcement-learning approach[J]. Computers and Electronics in Agriculture, 2025, 237: 110651.
[71]
FarooqS S, RahmanH, AbdulW S, et al. Curiosity-driven exploration in reinforcement learning: an adaptive self-supervised learning approach for playing action games[J]. Computers, 2025, 14(10): 434.
[72]
ShenC Q, MaoS H, XuB Z, et al. Spiral complete coverage path planning based on conformal slit mapping in multi-connected domains[J]. International Journal of Robotics Research, 2024, 43(14): 2183-2203.
[73]
LiuX, WenS, GuoZ, et al. Transfer learning-based sparse-reward meta-Q-learning algorithm for active SLAM[J]. Expert Systems with Applications, 2026, 299: 129799.
[74]
TobinJ, FongR, RayA, et al. Domain randomization for transferring deep neural networks from simulation to the real world[C]// IEEE/RSJ International Conference on Intelligent Robots and Systems. Vancouver: IEEE, 2017: 23-30.
[75]
FinnC, AbbeelP, LevineS. Model-agnostic meta-learning for fast adaptation of deep networks[C]// Proceedings of the 34th International Conference on Machine Learning. Sydney: PMLR, 2017: 1126-1135.
[76]
PanS J, YangQ. A survey on transfer learning[J]. IEEE Transa-ctions on knowledge and data engineering, 2010, 22(10): 1345-1359.
ZhuangFuzhen, LuoPing, HeQing, et al. Research progress of transfer learning[J]. Journal of Software, 2015, 26(1): 26-39.
[79]
XuX, LiuX, ZhaoY, et al. Efficient dynamic path planning algorithm for asvs based on transfer reinforcement learning[J]. IEEE Transactions on Intelligent Transportation Systems, 2025, 26(11): 19270-19280.
LiangJunjie, WeiJianjing, JiangZhengfeng. A survey of generative adversarial networks (GAN)[J]. Journal of Frontiers of Computer Science and Technology, 2020, 14(1): 1-17.
[85]
QuangT V, DoanD T. Online transfer learning(OTL) for accele-rating deep reinforcement learning (DRL) for building energy management[J]. Journal of Building Performance Simulation, 2025.DOI: 10.1080/19401493.2025.2511826.
[86]
XuC, InuiguchiM, HayashiN, et al. Enhancing reinforcement learning-based energy management through transfer learning with load and PV forecasting[J]. IEEE Access, 2025, 13: 43956-43972.
[87]
ZhaoG, ChengY, JiaY, et al. A transfer-learning-based STL-LSTM framework for significant wave height forecasting[J]. Journal of Marine Science and Engineering, 2026, 14(2): 146.
[88]
RamosD, LacasaL, ValeroE, et al. Transfer learning-enhanced deep reinforcement learning for aerodynamic airfoil optimization subject to structural constraints[J]. Physics of Fluids, 2025, 37: 85194.
GuoShujie, TianHua, WangWei. A solution method for path planning problems in complex scenarios[J]. Computer Technology and Development, 2022, 32(3): 27-33.
YuXinyi, LiuFei, OuLinlin. Multi-agent path planning method based on graph neural networks[J]. Chinese High Technology Letters, 2024, 34(10): 1081-1090.
[93]
ShahedS, ArmanS M, SamiS H, et al. A hybrid approach for facial parsing using transfer learning[J]. Scientific Reports, 2026, 16: 3405.
XingQiang, ChenZhong, LengZhaoying, et al. Path planning and charging navigation strategy for electric vehicles based on real-time traffic information[J]. Proceedings of the CSEE, 2020, 40(2): 534-550.
DouMingxing. Key technology research on autonomous path planning of reconfigurable robots for cross-floor scenarios[D]. Chongqing: Chongqing University, 2023.
LiYongdi. Research on local path planning of mobile robots based on deep reinforcement learning under ROS[D]. Zibo: Shandong University of Technology, 2023.
[100]
FarooqF, AliZ A, ShafiqM, et al. Intelligent planning of UAV flocks via transfer learning and multi-objective optimization[J]. Arabian Journal for Science and Engineering, 2025, 50: 16089-16106.
[101]
SongY, XuY, PiD, et al. Competitive many-task differential evolution with reinforcement learning and meta-knowledge transfer[J]. Knowledge-Based Systems, 2025, 326: 113931.
[102]
HuH, LiuZ, WuF. RL-PMO: a reinforcement learning-based optimization algorithm for parallel SFC migration[J]. Sensors, 2026, 26(1): 242.
[103]
ZhuZ, HanY, DuA. Enhanced eco-driving and energy management for heterogeneous hybrid electric vehicles through integrated transfer learning and deep reinforcement learning[J]. Energy, 2026, 342: 139626.
[104]
CarvalhoJ P, AguiarA P. Multi-agent reinforcement learning for zero-shot coverage path planning with dynamic UAV networks[J]. Robotics and Autonomous Systems, 2026, 195: 105163.
[105]
AljabriM A, AjourM N, BahabriM O, et al. Hierarchical deep reinforcement learning and model predictive control for voltage-aware electric vehicle charging coordination in distribution network[J]. IEEE Access, 2026, 14: 1123-1140.