Self-Organizing Traffic Lights as an Upper Bound Estimate

Self-organizing traffic lights (SOTL) methodology is highly popular and is the subject of much discussion (e.g., [1–7]). Zhung et al. [1] (based on constructions presented in [2]) claim that by using SOTL “overall better performance and higher capacity can be achieved” in comparison to the systems normally used by traffic authorities. “The idealized control scheme SOTL, which is designed to uniformize the network density distribution, always results in a higher MFD compared to the commonly used SCATS system. SOTL increases the network capacity and produces higher flows in the congested regime.” [1, Section 7]. In contrast, our analysis of the methodology and results in [1] has shown that, although SOTL provides a way of building the plot of an upper bound estimate for flow, this upper bound estimate may not be realistic. The corresponding traffic control schedules may in fact be impractical and unachievable. The practical schedule (i.e., the best solution for real traffic control operations) could have its plot slightly below the SOTL-based plot but be realistic and achievable. Hence, the previous statements about SOTL outperforming the Sydney Coordinated Adaptive Traffic System (SCATS) are not correct. The statement in [1] that the “upper bound estimate plot based on SOTL is higher than the plot related to the particular model of one commonly used traffic control system” is not supported by evidence. The purpose of this paper is neither a full-scale analysis of [1] nor the search for mathematical or logical inconsistencies in it. It is just a demonstration that the SOTL model [1] can provide upper bound esti-