Logistic routing problems inherently include mathematical difficulties in solution optimization and also strongly depend on specific load distributions in path network. In this study, to overcome these difficulties, we introduce artificial potentials in network topology in an environment where uncollected and undelivered goods are located in nodes with goods currently carried by agents. This inclusion of the hypothetical physics-informed effects in our reinforcement model enhances rapid convergence of logistic route solution, whose optimization level is confirmed by higher-order Markov chain networks. This method also works in cases where a network includes edges with different weights like highway effects, close to specific calculation environments along with real-world demands.