Influence of meta-heuristic optimization on the performance of adaptive interval type2-fuzzy traffic signal controllers

Intelligent traffic control systems optimized using meta-heuristic algorithms can greatly alleviate traffic congestions in urban areas. Meta-heuristics are broadly used as efficient approaches for complex optimization problems. Comparing the performance of optimization methods on different applications is a way to evaluate their effectiveness. The current literature lacks studies on how performance of traffic signal controllers is affected by utilized optimization algorithms. This paper evaluates the performance of three meta-heuristic optimization methods on an advanced interval type-2 adaptive neuro-fuzzy inference system (IT2ANFIS)-based controller for complex road networks. Simulated annealing (SA), genetic algorithm (GA), and the cuckoo search (CS) are applied for optimal tuning of IT2ANFIS controller. Optimizations methods adjust the parameters in a way to reduce the total travel time of vehicles in the road network. Paramics is used to design and simulate urban traffic network models and implement proposed timing controllers. Comprehensive simulation and performance evaluation are done for both single and multi-intersection traffic networks. Obtained results reveal significant superiority of IT2ANFIS trained using CS method over other controllers. The average performance of the CS-IT2ANFIS is about 31% better than the benchmark fixed-time controllers. This is 17% and only 3% for GA-IT2ANFIS and SA-IT2ANFIS controllers respectively.