CityTrajBench: A Unified Benchmark for City-Scale Vehicle Trajectory Generation

• URL: http://arxiv.org/abs/2606.02287v1
• Date: Mon, 01 Jun 2026 14:08:56 GMT
• Title: CityTrajBench: A Unified Benchmark for City-Scale Vehicle Trajectory Generation
• Abstract summary: CityTrajBench is a unified benchmark framework and protocol for city-scale vehicle trajectory generation.<n>It standardizes data ingestion, trajectory normalization, feature construction, model adaptation, map-aware post-processing, model selection, and evaluation.<n>The benchmark measures global spatial realism, trip-level distribution fidelity, trajectory-level geometric similarity, conditional mobility consistency, and efficiency.
• Score: 11.800237669997472
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Urban trajectory generation is a fundamental task for transportation simulation, urban planning, and mobility analytics. However, systematic comparison across trajectory generation methods remains difficult because existing studies often rely on different datasets, preprocessing pipelines, trajectory representations, and evaluation metrics. This fragmentation makes it unclear whether reported performance differences arise from the generation mechanism itself or from inconsistent experimental protocols. To address this issue, we present CityTrajBench, a unified benchmark framework and protocol for city-scale vehicle trajectory generation. CityTrajBench standardizes data ingestion, trajectory normalization, feature construction, model adaptation, map-aware post-processing, model selection, and multi-level evaluation under a common setting. It supports heterogeneous generators, including statistical baselines, VAE-based, GAN-based, diffusion-based, and flow-matching-based models, and evaluates them on three real-world urban trajectory datasets. The benchmark measures global spatial realism, trip-level distribution fidelity, trajectory-level geometric similarity, conditional mobility consistency, and efficiency. Experiments reveal clear trade-offs across model families: DiffTraj is strongest on trajectory-level geometric fidelity, DiffRNTraj is competitive on structure-sensitive global realism, and TrajFlow provides a strong balance across realism, quality, conditional consistency, and efficiency. Meanwhile, a simple Markov baseline remains competitive on coarse-grained trip and local-movement statistics. These findings show that urban trajectory generation quality is inherently multi-objective, that no single model dominates all criteria equally, and that CityTrajBench provides a reproducible benchmark protocol and testbed for future research on urban mobility generation.

Related papers

• CityGen: Structure-Guided City-Style Synthesis for Cross-City Autonomous Driving [5.149645190293104]
  CityTransfer-Bench is a benchmark for evaluating cross-city generalization across perception, segmentation, and planning.<n>CityGen is a diffusion-based generative framework that performs zero-label city adaptation via HD-map-conditioned synthesis guided by city-level visual prompts.
  arXiv  Detail & Related papers  (2026-05-28T13:46:48Z)
• Spatiotemporal Heterogeneity of AI-Driven Traffic Flow Patterns and Land Use Interaction: A GeoAI-Based Analysis of Multimodal Urban Mobility [0.0]
  Urban traffic flow governed by the complex nonlinear interaction between land use and mobility demand.<n>This study proposes a GeoAI Hybrid framework to model traffic flow patterns and their interaction with land use across three mobility modes.<n>Applying the framework to a dataset of 350 traffic analysis zones across six cities spanning two contrasting urban morphologies, four findings emerge.
  arXiv  Detail & Related papers  (2026-03-05T18:45:44Z)
• TopoCurate:Modeling Interaction Topology for Tool-Use Agent Training [53.93696896939915]
  Training tool-use agents typically rely on Supervised Fine-Tuning (SFT) on successful trajectories and Reinforcement Learning (RL) on pass-rate-selected tasks.<n>We propose TopoCurate, an interaction-aware framework that projects multi-trial rollouts from the same task into a unified semantic quotient topology.<n>TopoCurate achieves consistent gains of 4.2% (SFT) and 6.9% (RL) over state-of-the-art baselines.
  arXiv  Detail & Related papers  (2026-03-02T10:38:54Z)
• ReeMark: Reeb Graphs for Simulating Patterns of Life in Spatiotemporal Trajectories [5.137749546327247]
  We introduce Markovian Reeb Graphs, a novel framework for simulating trajectories that preserve Patterns of Life (PoLs) learned from baseline data.<n>Our approach generates realistic trajectories that capture both consistency and variability in daily life.<n>These results position Markovian Reeb Graphs as a scalable framework for simulation with broad applicability across diverse urban environments.
  arXiv  Detail & Related papers  (2025-10-03T16:25:11Z)
• Collaborative Imputation of Urban Time Series through Cross-city Meta-learning [54.438991949772145]
  We propose a novel collaborative imputation paradigm leveraging meta-learned implicit neural representations (INRs)<n>We then introduce a cross-city collaborative learning scheme through model-agnostic meta learning.<n>Experiments on a diverse urban dataset from 20 global cities demonstrate our model's superior imputation performance and generalizability.
  arXiv  Detail & Related papers  (2025-01-20T07:12:40Z)
• UniTraj: Learning a Universal Trajectory Foundation Model from Billion-Scale Worldwide Traces [64.24594320103066]
  Building a universal trajectory foundation model is a promising solution to address the limitations of existing trajectory modeling approaches.<n>We introduce UniTraj, a Universal Trajectory foundation model that aims to address these limitations through three key innovations.<n>First, we construct WorldTrace, an unprecedented dataset of 2.45 million trajectories with billions of GPS points spanning 70 countries.
  arXiv  Detail & Related papers  (2024-11-06T12:06:43Z)
• Revisiting Synthetic Human Trajectories: Imitative Generation and Benchmarks Beyond Datasaurus [4.522142161017109]
  Human trajectory data plays a crucial role in various applications such as crowd management and epidemic prevention.<n> synthetic human trajectory data is generated to simulate as close as possible to real-world human trajectories.<n>We propose MIRAGE, a huMan-Imitative tRAjectory GenErative model designed as a neural Temporal Point Process.
  arXiv  Detail & Related papers  (2024-09-20T09:07:27Z)
• Physics-guided Active Sample Reweighting for Urban Flow Prediction [75.24539704456791]
  Urban flow prediction is a nuanced-temporal modeling that estimates the throughput of transportation services like buses, taxis and ride-driven models. Some recent prediction solutions bring remedies with the notion of physics-guided machine learning (PGML) We develop a atized physics-guided network (PN), and propose a data-aware framework Physics-guided Active Sample Reweighting (P-GASR)
  arXiv  Detail & Related papers  (2024-07-18T15:44:23Z)
• MobilityGPT: Enhanced Human Mobility Modeling with a GPT model [15.16172813601417]
  We reformat human mobility modeling as an autoregressive generation task, leveraging the Generative Pre-trained Transformer architecture.<n>We propose a geospatially-aware generative model, MobilityGPT, to ensure its controllable generation.<n> Experiments on real-world datasets demonstrate MobilityGPT's superior performance over state-of-the-art methods.
  arXiv  Detail & Related papers  (2024-02-05T18:22:21Z)
• Time-series Generation by Contrastive Imitation [87.51882102248395]
  We study a generative framework that seeks to combine the strengths of both: Motivated by a moment-matching objective to mitigate compounding error, we optimize a local (but forward-looking) transition policy. At inference, the learned policy serves as the generator for iterative sampling, and the learned energy serves as a trajectory-level measure for evaluating sample quality.
  arXiv  Detail & Related papers  (2023-11-02T16:45:25Z)
• Physical-Virtual Collaboration Modeling for Intra-and Inter-Station Metro Ridership Prediction [116.66657468425645]
  We propose a unified Physical-Virtual Collaboration Graph Network (PVCGN), which can effectively learn the complex ridership patterns from the tailor-designed graphs. Specifically, a physical graph is directly built based on the realistic topology of the studied metro system. A similarity graph and a correlation graph are built with virtual topologies under the guidance of the inter-station passenger flow similarity and correlation.
  arXiv  Detail & Related papers  (2020-01-14T16:47:54Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.