Bidirectional Q-learning for recycling path planning of used appliances under strong and weak constraints  

作　　者：Yang Qi Jinxin Cao Baijing Wu 

机构地区：[1]Institute of Transportation Engineering,Inner Mongolia University,Hohhot,010010,China [2]Inner Mongolia Academy of Science and Technology,Hohhot,010010,China [3]Institute of Electronics and Information Engineering,Lanzhou Jiaotong University,Lanzhou,730070,China

出　　处：《Communications in Transportation Research》2024年第1期451-460,共10页交通研究通讯(英文)

基　　金：National Natural Science Foundation of China(Grant Nos.72461028,71961024,72161032,72061028,and 71971022);Key Technology Research Plan of Inner Mongolia Autonomous Region(Grant No.2019GG287).

摘　　要：With the continuous innovation in household appliance technology and the improvement of living standards,the production of discarded household appliances has rapidly increased,making their recycling increasingly significant.Traditional path planning algorithms encounter difficulties in balancing efficiency and constraints in addressing the multi-objective,multi-constraint challenge posed by discarded household appliance recycling routes.To tackle this issue,this study introduces a bi-directional Q-learning-based path planning algorithm.By developing a bi-directional Q-learning mechanism and enhancing the initialization method of Q-learning,the algorithm aims to achieve efficient and effective optimization of discarded household appliance recycling routes.It implements bidirectional updates of the state-action value function from both the starting point and the target point.Additionally,a hierarchical reinforcement learning strategy and guided rewards are introduced to minimize blind exploration and expedite convergence.By decomposing complex recycling tasks into multiple sub-tasks and seeking paths with superior performance at each sub-task level,the initial exploratory blindness is reduced.To validate the efficacy of the proposed algorithm,gridbased modeling of real-world environments is utilized.Comparative experiments reveal significant improvements in iteration counts and path lengths,thereby validating its practical applicability in path planning for recycling initiatives.

关 键 词：Path planning Q-LEARNING Waste electrical recovery Reinforcement learning Reward function 

分 类 号：TN9[电子电信—信息与通信工程]