Optimality versus run time for isolated signalized intersections

Hale, Douglas A.; Park, Byungkyu Brian; Stevanovic, Aleksandar; Su, Peng

doi:10.1016/j.trc.2015.02.015

Cited by 14 publications

(13 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…According to the study results here, this would achieve an average of 33% improvement for the 72 core test cases. According to the prior study results [11], the outdated heuristics in today's products would likely achieve an average of 22-23% improvement in these same cases.…”

Section: Discussionmentioning

confidence: 97%

“…This experiment extended a recen optimization study [11] comparing efficien heuristic methods including a genetic a simulated annealing, Tabu search, and tradition The study tested 72 intersection scenarios widest possible variety of topographies and s documented in source [11]. Each scenario mo of traffic as recommended by the HCM [21].…”

Section: Experimental Designmentioning

confidence: 99%

“…However unlike SPSA, a wealth of GA-based signal timing research has tested various parameter values. For isolated signalized intersections, at least two sensitivity analysis efforts [2,11] In the optimization tests conducted for this study, mutation rate was the only internal GA parameter -apart from the ubiquitous number of generations -that was varied extensively. This was because in the authors' experience with isolated intersections, mutation rate had the most effective and predictable impact on the exploration-exploitation ratio.…”

Section: B Ga Internal Parametersmentioning

confidence: 99%

“…In the prior study [11] of signal timing at isolated intersections, signal timings and traffic demands for 72 test cases were pre-adjusted, so that overall intersection delays would always fall between 80 and 85 s/veh. This was a logical starting point for optimization because (1) intersections with delays exceeding 80 s/veh are considered to operate at a failing level of service [21], (2) congestion levels much higher or lower than this could reduce the amount of optimization improvement that can take place, and (3) severely oversaturated conditions are only analyzed accurately by timeconsuming, multi-period models.…”

Section: Starting Pointsmentioning

confidence: 99%

“…Genetic algorithms (GA) in particular have been rated favorably for signal optimization by several researchers [8,9], and have been commercialized within multiple desktop products. More recently, simulated annealing [10] was shown to optimize signal timings approximately as well as GA [11], at least for isolated intersections.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

An Investigation of SPSA for Signal Timing Optimization at Isolated Intersections

Hale

Antoniou

Park

et al. 2015

2015 IEEE 18th International Conference on Intelligent Transportation Systems

Self Cite

View full text Add to dashboard Cite

Simultaneous PerturbationStochastic Approximation (SPSA) has gradually gained favor as an efficient method for optimizing computationally expensive, "black box" traffic simulations. However, few recent studies have investigated the efficiency of SPSA for traffic signal timing optimization. It is important for this to be investigated, because significant room for improvement exists in the area of signal optimization. Some signal timing methods and products perform optimization very quickly, but deliver mediocre solutions. Other methods and products deliver high-quality solutions, but deliver those solutions very slowly. When using commercialized desktop signal timing products, engineers are often forced to choose between speed and solution quality. Real-time adaptive control products, which must optimize timings within seconds on a cycle-by-cycle basis, do not have enough time to reach a high-quality solution. Based on research results in the literature, SPSA holds the possibility of upgrading both desktop and adaptive solutions alike, by delivering high-quality solutions within seconds. This paper describes an extensive set of optimization tests involving SPSA. The final results suggest that today's signal timing solutions could be improved significantly by incorporating SPSA, genetic algorithms, and 'playbooks' of pre-optimized starting points.

show abstract

Section: Discussionmentioning

confidence: 97%

Section: Experimental Designmentioning

confidence: 99%

Section: B Ga Internal Parametersmentioning

confidence: 99%

Section: Starting Pointsmentioning

confidence: 99%