Evaluating Lateral Interactions of Motorized Two-Wheelers using Multi-Gene Symbolic Genetic Programming

Abstract

Complex maneuvering patterns are typical of motorized two-wheelers (MTWs), and their widespread adoption in many countries has spurred a growing response from transport researchers to model their dynamic behavior realistically. Considering the increased vulnerability of MTW drivers in dense urban mixed traffic systems, proper evaluation and modeling of lateral interactions between the drivers/riders moving abreast need to be addressed. A proper investigation can essentially help in understanding the behavioral aspects of riders in accepting shorter lateral clearances, design of exclusive motorcycle lanes, improved reliability of microsimulation models, and safety evaluation of the riders in a cognitive architecture. The current study therefore attempts to develop a novel symbolic regression model using a multigene genetic programming algorithm to generate and evaluate lateral clearance models naturally from field data for MTW interactions at mid-block sections, data being collected using video recorders. A binary logit model is initially developed to investigate the factors associated with the riders’ decisions to accept critical lateral clearances. Considering highly non-linear variations in data, the symbolic regression models were further developed and a comparison with the existing linear regression based lateral clearance models indicated that the symbolic model could generalize the non-linear structure of the data realistically and performed significantly better than the existing models.

Driving Behavior at Signalized Intersections Operating under Disordered Traffic Conditions

“…But the values in these studies are obtained from optimizing the macroscopic parameters of travel time or delay. It is well articulated that different variables or driving behavior representative parameters (DBRPs)-such as relative speed, spacing, acceleration of various vehicles under different conditions, space headways, lateral shift, and lane changes-form key inputs to different driving behavior models, especially for disordered traffic conditions (3,16,24). Therefore, the values of different DBRPs would govern these models' efficacy in modeling traffic behavior.…”

Model for Predicting Lateral Shifts and Driving Style of Motorized Two-Wheelers

“…They applied the Boruta algorithm to select potential features in the developed model, where only traffic-related features were considered and not the effect of the surrounding traffic environment. In contrast, recent studies by Das et al ( 15 ) and Das and Maurya ( 11 ) indicated the importance of considering the presence of leading and adjacent vehicles in estimating lateral clearance of MTWs while riding abreast. Besides, the execution of lateral shift is a process that depends on the number of attempts made by the rider to make a successful lateral shift, which indeed plays a significant role in understanding the behavioral intentions of a rider to make a shift and his/her driving style during the process.…”

“…Besides, the execution of lateral shift is a process that depends on the number of attempts made by the rider to make a successful lateral shift, which indeed plays a significant role in understanding the behavioral intentions of a rider to make a shift and his/her driving style during the process. Furthermore, with respect to methodological approaches, the binary logit model was one of the major modeling preferences adopted in several studies related to filtering and overtaking maneuvers of MTWs ( 8 , 11 , 15 ). However, some advanced models, such as multiple indicators–multiple causes (MIMIC) latent variable models ( 9 ), the decision tree ( 10 ), and the multigene genetic programming algorithm, were also applied to evaluate MTWs’ lateral interaction behaviors.…”

Stability Analysis of Mixed Traffic Flow using Car-Following Models on Trajectory Data

“…The driving behavior studies [2,3] from mixed traffic demonstrates smaller vehicles' lateral behavior impacting mixed traffic performance. Simultaneously, given the variation in the vehicles' physical properties, earlier studies [4,5] reported the underperformance of automated traffic tools in monitoring mixed traffic conditions. Due to the mixed traffic data constraints, the driving behavior studies from mixed traffic have not taken proper shape.…”