Calibration of SUMO for Indian Heterogeneous Traffic Conditions

Abstract

Efficient modelling of vehicular traffic is a challenging task in the context of Indian traffic conditions. One of the approaches for modelling traffic is using simulation. Though there are several traffic simulation software available, all of them are developed for the lane based and homogeneous traffic conditions. However, traffic conditions in many countries are heterogeneous and lane-less and for simulating such traffic, either specific software needs to be developed or calibration of existing software for such traffic conditions is required. For example, one of the commonly used software, namely VISSIM can be calibrated for such traffic conditions and is already reported in literature. However, VISSIM being licensed software, researchers have developed an open source software, namely Simulation of Urban MObility (SUMO). Though the initial development of SUMO focused on homogeneous and lane disciplined traffic, later researchers started developing modules for the Indian traffic with its wide mix of vehicle types and improper lane discipline. This paper presents a methodology for the calibration of SUMO for Indian heterogeneous traffic conditions by calibrating its parameters. Data from a 2 km segment in Chennai was used for the calibration. In the first level, parameters that can affect the driving behaviour under such conditions were identified using sensitivity analysis and one-way ANOVA test. Then optimal combination of parameters were identified using Genetic Algorithm (GA). Performance comparison was done with calibrated VISSIM for the same test bed. Average speed obtained from both the simulation software (VISSIM and SUMO) were compared and the errors were calculated in terms of Mean Absolute Percentage Error (MAPE) with respect to actual speed values. Results were found to be comparable, indicating that SUMO can be calibrated for simulating Indian traffic.