Key points are not available for this paper at this time.
Traffic congestion is an increasingly prevalent global issue that necessitates the advancement of Traffic Signal Control technologies. Deep Reinforcement Learning has emerged as a prominent machine learning paradigm, leveraging trial-and-error experimentation in conjunction with Deep Neural Network models to facilitate autonomous and coordinated management of traffic signal lights spanning numerous intersections within a traffic network. Reinforcement Learning methodologies employ diverse exploration strategies such as Epsilon greedy, Softmax, Upper Confidence Bound, among others, to ascertain an optimal policy. In the pursuit of long-term rewards, an effective exploration strategy must adeptly balance the exploitation of the current policy with the exploration of novel alternatives. The Epsilon greedy algorithm stands out as a widely adopted approach for navigating this trade-off in Reinforcement Learning. However, its performance is intricately tied to the initially handcrafted exploration rate. This work contributes significantly to the attainment of an optimal policy by primarily emphasizing two key aspects. Firstly, it underscores the criticality of meticulously tuning the -decay rate, which governs the progression of the exploration rate, in order to cultivate an optimal traffic signal control system. Secondly, this work delves into an in-depth exploration of the constraints inherent in the epsilon decay rate and offers potential avenues for future research in this domain.
Thadikamalla et al. (Thu,) studied this question.