MIL - The University of Tennessee
the university of tennessee machine intelligence lab

Intelligent Transportation Systems

Background and Motivation

Efficiently controlling the flow of traffic across complex transportation networks remains an intricate challenge. In recent work, the problem of scheduling traffic at an intersection has been addressed by structuring the problem as a Markov decision process (MDP). It has been shown that by using dynamic programming techniques, which aim to solve the Bellman equation given a stochastic model of the system, an optimal control strategy can be obtained. However, in real life, an accurate model of the system is not provided. Approximating a model yields limited results due to the inherent non-stationarity and non-Markovian properties of vehicular traffic flows. It is, therefore, important to develop techniques which not only improve the performance of the signaling system but also remain computationally modest.

Research Approach

We focus on developing, analyzing and evaluating algorithm designed for the signal control problem that employ concepts drawn from optimal control and computer networking. In addition to the classic performance metrics considered, such as delay and throughput, we also incorporate safety attributes into our algorithms. A theme of this study is establishing the stability characteristics of the proposed algorithms. Of particular interest are scenarios in which traffic is non-uniformly distributed across the lanes and is correlated traffic arrival patterns.