APPLICATION OF FUZZY LOGIC TO TRAFFIC SIGNAL CONTROL UNDER MIXED TRAFFIC CONDITIONS
Abstract
Traffic signal control is commonly used at road intersections to minimise vehicular \ndelay. Fixed time control shows good results in conditions where there is a little fluctuation in \ntraffic demand, however in time-varying traffic fixed time control becomes inflexible and \ninefficient. This may produce traffic congestion and lead to increased delays and air pollution. \nDemand responsive traffic signal control must be introduced to overcome these problems. \nHowever, all the available demand responsive traffic signal control methods such as \nVehicle Actuated Controller (VAC), Traffic Optimisation Logic (TOL), Microprocessor \nOptimised Vehicle Actuation (MOVA) and Fuzzy Logic Traffic Signal Controllers (FLTSC) have \nbeen developed for non-mixed traffic conditions, considering only motor vehicles move in \nclearly defined lanes, neglecting motorcycles. These demand responsive traffic signal controls \nare not appropriate for the mixed traffic conditions of developing countries such as Indonesia, \nwhere the traffic streams consist of different types of vehicle with a wide variation in their \nstatic, dynamic and operating characteristics, and with a particularly high proportion (30% - \n70%) of motorcycles. Also there is lack of lane discipline. \nThis thesis describes the design and evaluation of an adaptive traffic signal controller based on \nfuzzy logic for an isolated four-way intersection with specific reference to mixed traffic in \ndeveloping countries, including a high proportion of motorcycles. Four proposed controllers \nhave been developed for different schemes. The controllers were designed to be responsive to \nreal time traffic demands. The study identifies two traffic parameters as appropriate as input \ndata for an adaptive traffic signal controller under mixed traffic conditions such as the proposed \nFLTSC: the average occupancy rate (%) and maximum queue length (metres). The literature \nstudy suggest that this data should be collected using advances video image processing. The \nproposed FLTSC uses maximum queue lengths and average occupancy rates collected during the \nprevious cycle to estimate the number of seconds of green time required by each set of signal \ngroups during the next cycle. \nThe effectiveness of the proposed FLTSC was analysed using the microscopic traffic \nsimulation model VISSIM. Prior to doing so, the VISSIM model was calibrated and validated. \nFrom the validation process it was apparent that the VISSIM model could be adapted to simulate mixed traffic conditions by use of the Packet approach. In this approach, motorcycles \nare modelled as a group of motorcycles. \nThe performance of the proposed FLTSC was contrasted with a Fixed Time Controller \n(FTC) for different case studies on a simulated four-way intersection. The FTC is represented by \nthe calculation as suggested in the Indonesian Highway Capacity Manual. Separate analysis \nusing TRANSYT show that this is a valid assumption to make. The simulation results show that \nthe proposed FLTSC is generally better than the FTC in terms of the average delay of vehicles at \nan intersection, especially under time-varying traffic. \nFurther analysis was carried out to compare the performance of the proposed FLTSC \nagainst a Vehicle Actuated Controller (VAC) for different traffic conditions on a simulated four- \nway intersection, East-West and North-South without turning movements. In order to analyse \nthe performance of VAC, a refined VISSIM model was developed. This used the latest version of \nthe VISSIM software and allowed individual vehicles (and particularly motorcycles) to be \nmodelled in mixed traffic. \nThe phase extension time is one of the most critical parameters to affect the overall \nperformance of VAC (Bullen, 1989). To provide a fair comparison of the performance between \nthe proposed FLTSC and the VAC, an investigation was carried out to find the most appropriate \nextension time for the VAC that was suitable for mixed traffic. The effect of motorcycles to the \nperformance of the VAC was also investigated. Two schemes were carried out to observe it, \nnamely: Scheme 1 where detector detects all vehicle types (DfT, 2006) and Scheme 2 where \ndetector detects all vehicle types, apart from motorcycles. \nThe simulation results show that the VAC System D (DfT, 2006) using an extension time \nof 1.2 seconds and the VAC Extension Principle (Kell and Fullerton, 1991) with a detector \nposition of 30 metres and extension time of 3.0 seconds produced better performance than the \nother extension times tested for both schemes in terms of the average delay of vehicles. This is \nslightly shorter than current practice in developed countries. \nThe simulation results indicate that the performance of the VACs with scheme 1 is \ngenerally worse than with scheme 2. The performance of the VACs with scheme 1 against \nscheme 2 tended to reduce significantly as the percentage of motorcycles in traffic increased. \nThe study compares the effectiveness of FTC, VAC Extension Principle (VAC-EP), VAC System \nD (VAC-SD) and proposed FLTSC in various traffic conditions. The simulation results indicate \nthat the average delay of the proposed FLTSC is close to the average delay of the FTC when used \nin cases with constant traffic flows but sometimes worse. However, in cases of time-varying \ntraffic the proposed FLTSC is superior to the FTC. When comparing the simulation results of the \nproposed FLTSC, VAC-SD and VAC-EP, again the proposed FLTSC does not improve average \ndelay, when traffic flows constant but produces better results in cases of time-varying traffic.