Agent-Based Modeling for Predicting Pedestrian Trajectories Around an Autonomous Vehicle

Abstract

This paper addresses modeling and simulating pedestrian trajectories when interacting with an autonomous vehicle in a shared space. Most pedestrian–vehicle interaction models are not suitable for predicting individual trajectories. Data-driven models yield accurate predictions but lack generalizability to new scenarios, usually do not run in real time and produce results that are poorly explainable. Current expert models do not deal with the diversity of possible pedestrian interactions with the vehicle in a shared space and lack microscopic validation. We propose an expert pedestrian model that combines the social force model and a new decision model for anticipating pedestrian–vehicle interactions. The proposed model integrates different observed pedestrian behaviors, as well as the behaviors of the social groups of pedestrians, in diverse interaction scenarios with a car. We calibrate the model by fitting the parameters values on a training set. We validate the model and evaluate its predictive potential through qualitative and quantitative comparisons with ground truth trajectories. The proposed model reproduces observed behaviors that have not been replicated by the social force model and outperforms the social force model at predicting pedestrian behavior around the vehicle on the used dataset. The model generates explainable and real-time trajectory predictions. Additional evaluation on a new dataset shows that the model generalizes well to new scenarios and can be applied to an autonomous vehicle embedded prediction.

Pedestrian Trajectory Prediction in Pedestrian-Vehicle Mixed Environments: A Systematic Review

“…In a couple of expert-based models (e.g., SFM), the parameters of the model are estimated by using real-world trajectory data [17], [19], [23], [68], [69]. In these methods, parameters such as safety distances, range of repulsive forces, payoff matrices of the game [19] or the social factor and perception zones in the SFM [68] are calibrated through a process to get the minimum difference between the ground-truth trajectories in the dataset and the simulated ones using the model.…”

“…From a broad perspective, these interaction effects can be categorized to be modelled either explicitly or implicitly. In the explicit interaction modelling approaches (e.g., see [17], [18], [70], [71]), the effect of the vehicle on a pedestrian's motion is forced through some clear terms in the formulation of the pedestrian's motion such as through explicit forces in the social force model (e.g., [68], [72]). On the other hand, datadriven modules often account for these interactions implicitly by using the vehicle's trajectory as another input to the model along with the target pedestrian's own trajectory (e.g., see [30], [45]).…”

“…Some papers have modelled the vehicle with a circular shape same as how pedestrians are modelled in SFM but with a larger radius [68], [77]. In these models, the asymmetrical effect of the vehicle's influence is accounted for in the anisotropy function or considering the approach angle and its change rate [68], [77].…”

Driver-pedestrian interactions at unsignalised crossings are not in line with the Nash equilibrium

“…On the other hand, glass box models offer the advantage of interpretability and transparency by providing explanations for the mechanisms in relatively great detail. These models rely on different modelling paradigms including agent-based modelling (Bonabeau, 2002;Prédhumeau et al, 2022), optimal control theory (Le & Malikopoulos, 2022;Ross, 2015), Markovian processes (Bellman, 1957;Hsu et al, 2018), evidence accumulation (Pekkanen et al, 2022;Ratcliff et al, 2016), proxemics (Domeyer et al, 2019), discrete choice modelling (Hensher & Johnson, 2018;Zhao et al, 2019) and game theory (Elvik, 2014).…”

How Do Human-Driven Vehicles Avoid Pedestrians in Interactive Environments? A Naturalistic Driving Study

“…The deployment of autonomous vehicles (AVs) is still facing many challenges, particularly their application in complex interactive environments, where different types of road users are included, e.g., pedestrians [ 1 , 2 , 3 ]. In such interactive environments, AVs not only need to make appropriate motion decisions but also need to consider other road users’ responses to AVs [ 4 ]. However, currently, AVs generally consider pedestrians as dynamic obstacles without intention and generally make simple braking decisions and actions to stop when AVs encounter pedestrians, especially in the scenarios of pedestrians crossing the street [ 5 ].…”