The diverse nature of vehicle categories and the resultant lane discipline in mixed (heterogeneous) traffic cause complex spatial interactions. As a result, the driving behavior process in mixed traffic conditions is meaningfully different, where both longitudinal and lateral movements of the vehicles continuously occur. Under prevailing homogeneous traffic conditions in developed countries, driving behavior is partially discrete, where following longitudinal behavior and outboard lane-change models can model traffic behavior. However, the established car-following and lane-change models cannot be directly used in shaping mixed traffic conditions. Such conditions also warrant the use of high-quality microlevel vehicular trajectory data. Accordingly, realizing this need, vehicular trajectory data for different traffic flow conditions were developed. The data were used to extract the parameters required for modeling the vehicles' positions using machine learning algorithms. Three established supervised machine learning algorithms (k-NN, random forest, and regression tree) and deep learning are selected to model mixed traffic conditions. The parameters which influence longitudinal and lateral movements are identified using Spearman correlation analysis. Furthermore, simulation runs are performed using the python language. The performance of the algorithms is evaluated both at the microscopic and macroscopic levels using relevant traffic indicators. The results show that a deep learning model and k-NN tend to replicate better-mixed traffic conditions than random forest and regression trees.

The present work introduced a framework of developing comprehensive extended vehicular trajectory data under heterogeneous non-lane-based traffic conditions like the NGSIM datasets in the United States. Due to the absence of automation and instrumentation, and even the lack of sensor deployment on roads in developing economies like India, it is even more challenging to study driver behavior. A new stitching-based algorithm was used for developing the extended trajectory database for three traffic-flow levels on a 535-m long section of an urban arterial. The algorithm was used to stitch the trajectory data over the segments such that the subject vehicle with continuous trajectory data points over the entire study stretch. The developed framework is a novel tool for establishing a trajectory dataset for mixed traffic, it should be of interest to researchers in developing and developed countries.

The present research work provides a methodology of calibration of the vehicle-following behavior in VISSIM using vehicle trajectory data obtained under non-lane-based mixed traffic conditions. In this context, two access-controlled mid-block sections, one on the multi-lane urban road with six-lane-divided carriageway and the second one with eight-lane-divided carriageway are identified for extracting high-quality vehicle trajectory data. Later, using trajectory data, lane-wise time-space plots were developed for the vehicles’ movement simultaneously in the same lane as well as from adjacent lanes for studying non-lane-based movement.The investigation is carried out by developing a relationship (hysteresis plots) between relative distance versus relative velocity among the leader–follower vehicle for the vehicles in the same as well as in adjacent lane that is nearby position. Hysteresis phenomenon for vehicles under the following behavior is examined.Data about identified potential vehicular pairs are used for calibration of well-known psychophysical car-following models.

The research work presented is originated with the intent of studying driving behavior, especially in a lateral direction under the influence of construction work activities under prevailing heterogeneous traffic conditions. Initially, a midblock road section is selected, and then macroscopic traffic characteristics were evaluated. Later, Mumbai metro-rail construction work was started on the same road section. As a result, it was planned to study the variation in traffic-flow characteristics for road sections with-and-without construction work zone at microscopic as well as macroscopic levels. For this purpose, video graphic surveys were conducted, and macroscopic characteristics were evaluated for both study sections (with-and-without construction work zone). The results showed that at given flow levels, the stream speeds were dropped from 70 to 50 kph, followed by a drop in efficiency in terms of per-lane capacity. To sense this impact more deeply, vehicular trajectory data were developed over road sections at three different traffic-flow levels. The driving behavior is investigated using different variables, such as longitudinal following behavior, following times, lateral amplitude, and lateral placement, followed by lateral direction correlation analysis. From the investigation, the results show that the following-hysteresis plots for vehicles are found to be slender in a construction work-zone, revealing the conservative approach in driving behavior. Furthermore, following-time analysis also supports the same interpretation. Given the smaller size of vehicles, the vehicle tends to possess extra lateral freedom and hence is responsible for the seeping phenomenon through a porous media in the traffic stream. As an impact of reduced road width, capacity (per-lane) values are found to be higher in the base section than the section having a construction work zone. Further, based on the conceptualized framework, safety in the traffic stream over the study sections was evaluated. Finally, the importance of cautioning drivers the construction work activity in advance was well revealed from the safety analysis of this work.

The present research work is initiated, probing for safety levels, evolving as an outcome of analysis of micro-level interactions due to naturalistic vehicular movements on Indian expressways. The research work reported here, devised a micro-level parameter, namely instant perception time (IPT) as a safety measure, which is a function of vehicular interactions. As a step towards deriving IPT, video-graphic surveys were conducted and traffic data was collected on two intercity Indian expressways. As a part of this work, first microscopic trajectory data was developed for capturing naturalistic driving behavior under heterogeneous traffic environment. Thereafter, hysteresis plots were developed between leader-follower pairs using relative speed and spacing between the vehicles. By means of the developed hysteresis plots, a novel approach for identification and quantifying chances of rear-end crashes using follower's Instantaneous perception time (IPT), is proposed in the present study. To study the IPT approach under different flow conditions, well-calibrated and validated simulation model is also developed using real trajectory data collected for the prevailing traffic conditions observed on the study sections. From the obtained IPTs using simulation model, three cluster ranges are delineated pertaining to different levels of risks of crashes. The cluster-analysis suggests that the least values of space-headway (m) and IPT (s) (space headway < 10.91 m, IPT < 5.61 s) at a given location shows most severe chances of rear-end collision. While, lower to moderate values indicate (with thresholds: space headway in the range 10.91–18.5 m; IPT 5.61–11.91 s), whereas higher values of IPT (more than 12 s) suggest the least chances of rear-end collision. The methodological approach, results and findings thus developed under naturalistic driving environment, presented here can be very useful in understanding the operational deficiencies of traffic streams, road geometry, coupled with safety assessment of traffic streams on uninterrupted roadway facilities. Furthermore, traffic control or geometry improvement related decisions can also be taken at different unsafe locations using IPT as a risk measure to enhance safety.

This research work is focused on modeling vehicle following behavior under mixed traffic conditions. Vehicle-following behavior can be potentially utilized in building real-time drivers’ assistance systems or identifying collision escaping time thresholds for varying roadway and traffic conditions. For this purpose, initially, vehicular trajectory data was developed over the road sections in India. Thereafter, based on hysteresis phenomenon among the vehicles, vehicles in the following conditions are identified. Further to model the following the behavior of vehicles, selected car-following models, such as Wiedemann-74, Wiedemann-99, Gipps, Bando and Intelligent Driver Model (IDM) are calibrated for followers (subject vehicle) from each vehicle category. The model’s validity is tested by comparing the observed position and modeled position of the follower under varying conditions. Subsequently, to test the car-following model’s efficacy in replicating the following behavior, simulation runs were performed using VISSIM 9.0. For this purpose, the car following models (other than Wiedemann models) were coded in VISSIM using external driving behavior program. The models were then tested using calibrated vehicle-dependent following-behavior parameters. Based on the simulation runs, the comparison was made with derived traffic parameters for macroscopic validation. For better understanding hysteresis plots were also compared for different vehicle-following models, using simulated trajectory data. Further, simulated hysteresis for different vehicle types was compared with related plots made using actual trajectory data. It was well-established that psychophysical models (Wiedemann), multi-regime (Gipps), and IDM car-following models resemble Indian traffic behavior reasonably well, whereas single-regime car-following model, optimal velocity model fails in replicating the actual following-behavior.