Realistic Data-Driven Traffic Flow Animation Using Texture Synthesis

Abstract

We present a novel data-driven approach to populate virtual road networks with realistic traffic flows. Specifically, given a limited set of vehicle trajectories as the input samples, our approach first synthesizes a large set of vehicle trajectories. By taking the spatio-temporal information of traffic flows as a 2D texture, the generation of new traffic flows can be formulated as a texture synthesis process, which is solved by minimizing a newly developed traffic texture energy. The synthesized output captures the spatio-temporal dynamics of the input traffic flows, and the vehicle interactions in it strictly follow traffic rules. After that, we position the synthesized vehicle trajectory data to virtual road networks using a cage-based registration scheme, where a few traffic-specific constraints are enforced to maintain each vehicle's original spatial location and synchronize its motion in concert with its neighboring vehicles. Our approach is intuitive to control and scalable to the complexity of virtual road networks. We validated our approach through many experiments and paired comparison user studies.

A Survey on Visual Traffic Simulation: Models, Evaluations, and Applications in Autonomous Driving

“…The generated traffic flows via various traffic simulation and animation approaches: (a) synthesized traffic flows on a highway network from Chao et al. [CDR*18], (b) a dense traffic scenario with signalized crossing from Shen and Jin [SJ12], (c) the reconstruction of virtual traffic flows using in‐road sensor data from Wilkie et al. [WSL13], (d) the reconstructed city‐scale traffic using GPS data from Li et al.…”

“…Chao et al . [CDR*18] synthesize new vehicle trajectories through a fusion of texture synthesis and traffic behaviour rules, using a limited set of vehicle trajectories as input samples. The example (input) vehicle trajectory set contains a variety of traffic flow segments in terms of the number of lanes and flow density.…”

“…Texture analogy of a set of two‐lanes vehicle trajectories [CDR*18]. The spatial‐temporal information of the trajectory set can be conceptually viewed as a 2D texture, and each traffic texel encodes a vehicle's states at a certain frame, including its movement information and position relationship with its neighbouring vehicles.…”

“…[CDR*18], researchers conducted user studies [KS40] using pairwise comparison on the generated traffic flows with three different methods: (1) the ground‐truth (i.e. the NGSIM traffic flow data), (2) the proposed texture‐based traffic synthesis method [CDR*18] and (3) one of the latest developments of the IDM model [SJ12]. For each test scene, three different traffic flow animations are generated using the above three different approaches, respectively.…”

“…This phenomenon gives rise to data‐driven traffic animation techniques. Example works include the reconstruction of traffic flows from spatio‐temporal data acquired by existing in‐road sensors [SVDBLM11, WSL13, LWL17], the synthesis of new traffic flows from limited trajectory samples [CDR*18] and the generation of traffic flows through learning behaviour patterns and individual characteristics from traffic monitoring data sets [CSJ13, BMWD16].…”