ABSTRACT In the university mental health management, counsellors are required to conduct the periodic follow‐up visits to students who have been identified as special concern. However, how to plan patrol routes and determine visit frequency for each student in a reasonable way is still a challenging problem in practice. This paper proposes a state‐driven periodic patrol routing model (SD‐PPRM), which incorporates Markov transition probabilities into periodic vehicle routing framework. In this model, the visit frequency of each student is no longer fixed at the initial assessment but can be adjusted adaptively according to the changes of psychological state after each visit. Meanwhile, covering constraints are introduced to make sure that all registered students can be reached by the patrol routes. For solving this problem, a reinforcement‐learning‐based method is designed, which adopts proximal policy optimization algorithm with attention‐based encoder and nearest‐neighbour/2‐opt route construction procedure to learn the patrol policies through simulated interaction with stochastic state dynamics. The computational experiments are carried out on the simulated campus instances with 15, 30 and 50 students. The results show that SD‐PPRM can reduce the proportion of students who remain in Critical state compared with static and greedy baselines, and at the same time, it maintains a competitive routing efficiency and achieves coverage rates above 90% under a coverage radius of 200 m. In addition, the sensitivity analyses are conducted to verify the robustness of proposed model under different parameter settings including state penalty weight, coverage radius and intervention strength.
J et al. (Tue,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: