Enhanced social force model for non-motorized vehicle flow: predictive decision-making and spatio-temporal repulsion mechanism

Abstract

Non-motorized vehicles account for a growing share of urban travel, and interactions in non-motorized lanes are increasingly complex. Most microscopic models, such as the social force model, rely on instantaneous distances and thus fail to capture anticipatory decision-making and time-dependent interactions, limiting their ability to reproduce proactive behaviors observed in real traffic. To address these shortcomings, this study proposes an enhanced social force model for non-motorized vehicles to simulate their complex interactive behaviors on real road segments, encompassing four primary behaviors: free movement, overtaking, avoiding, and following. The model incorporates a predictive mechanism based on current speed and position, enabling individuals to autonomously decide subsequent actions based on future states. It also refines the traditional repulsive force calculation by integrating contact duration between non-motorized vehicles into the force expression. By combining this with techniques such as altering the direction of the driving force, the model enables proactive overtaking of slower vehicles ahead and early avoidance of wrong-way traffic, making the non-motorized cycling process more closely resemble real-world dynamic changes. The improved model was validated using real-world data collected in Nanjing, China. The results demonstrate that this model can accurately reproduce the actual movement characteristics of non-motorized vehicles. It also reduces potential conflicts in non-motorized vehicle lanes, enhances lane flow efficiency, and exhibits higher simulation accuracy compared to the original social force model.