Interaction-Aware Motion Prediction at Highways: A Comparison of Three Lane Changing Models

Abstract

The behavior of traffic participants is full of uncertainties in the real world. It depends on their intentions, the road layout, and the interaction between them. Probabilistic intention and motion predictions are unavoidable to safely navigate in complex scenarios. In this work, we propose a framework to compute the motion prediction of the surrounding vehicles taking into account all possible routes obtained from a given map. To that end, a Dynamic Bayesian Network is used to model the problem and a particle filter is applied to infer the probability of being on a specific route and the intention to change lanes. Our framework, based on Markov chains, is generic and can handle various road layouts and any number of vehicles. The framework is evaluated in two scenarios: a two-lane highway and a three-lane merging highway. Finally, the influence of a set of lane-changing methods is evaluated on the predictions of the vehicles present on the scene.

GPU-Accelerated Interaction-Aware Motion Prediction

“…These interactions are computed for every vehicle in the scene against the rest and take into account both information that is coded into the traffic rules and also potential deviations from the expected road-safe behavior to deal with unsafe driving situations, such as violating a stop line. This feature makes this interaction-aware motion prediction algorithm especially useful in complex situations with dense traffic, like unsignalized intersections [20], sudden unsafe lane changes [21], or roundabouts [22]. The intentions are then fused with the motion predictions computed with a kinematic model to produce a 3D motion grid used by the ego vehicle to navigate through the scene.…”

“…The main parts where the GPU-based acceleration has significantly reduced the computation time are highlighted in blue in the flowchart, namely in the particle filter used to compute the intentions and in the matrix multiplication of the motion prediction. This flowchart is extensively described in [20][21][22], where the algorithm is validated and compared with state-of-the-art approaches. Hence, it will only be briefly reviewed below.…”

“…• The relations obtained from the pairwise analysis of the corridors are used to estimate the longitudinal E s [20] and lateral E c [21] behaviors of the vehicles.…”