Understanding how many aircraft, and which traffic configurations, can be safely and efficiently managed by air traffic controllers within a given airspace is fundamental to air traffic management. Air traffic controllers continuously modify aircraft trajectories through interventions such as speed advisories and radar vectoring, and the patterns of these interventions are difficult to capture effectively using existing air traffic simulators. To address this limitation, we propose Diffusion Transformers for Traffic Scene Generation (DiTSG), which learns to generate air traffic scenes from recorded multi-aircraft trajectories. DiTSG employs stacked multi-agent inverted Transformer layers as the denoising backbone of a diffusion transformer, jointly generates the trajectories of multiple aircraft within a traffic scene, and conditions the generation on agent-specific required times of arrival. The resulting trajectories generated by DiTSG preserve realistic inter-aircraft interactions observed in actual air traffic operations. Experiments show that the DiTSG generates diverse traffic scenes that reproduce the statistical properties of recorded traffic while preserving realistic inter-aircraft interactions.
Figures & tables
Fig. 1: Overview of DiTSG . Left: a DiTSG block, which applies masked multivariate attention, agent attention, and a feed-forward layer, each modulated by the conditioning vector Y through adaLN. Right: the sampling loop, which tokenizes the noisy scene Xs , passes it through L blocks and a linear layer to obtain the velocity Vθ , and integrates the ODE for K Euler steps from noise to a traffic scene.
Method
Docc
Rnov
Diversity
SVR (%)
Real
–
1.000
–
0.000
MA-VAE
0.033
0.947
16.690
0.250
DiTSG
0.015
0.953
14.938
0.101
TABLE I: Evaluation of generated traffic scenes.
Fig. 2: Separation patterns in real and generated traffic. Distribution of the minimum horizontal (left) and vertical (right) separation between aircraft pairs. Dashed lines indicate the separation minima.
Fig. 3: Traffic fidelity across arrival stages and diffusion timesteps. JSD between real and generated traffic within each RTA ( τ ) range (left). Horizontal and vertical mean absolute error between DiTSG’s predicted and true positions as a function of the diffusion timestep s (right).
Fig. 4: Generated traffic scenes with varying numbers of aircraft. Examples of scenes generated by DiTSG for N=6 to N=11 . Each panel shows one scene over the ICN terminal area, with STARs and runways shown in gray for reference. Aircraft are labeled AC1 to AC N in landing order, and each label reports the RTA τ , current altitude, and ground speed.
Fig. 5: Generated alternatives under the same RTA condition. Comparison between a recorded traffic scene and the closest scenes generated by DiTSG under the same τ .
Fig. 6: Agent attention throughout the denoising process. The query aircraft is shown in orange, while the other aircraft are colored according to the attention they receive from the query. Intermediate states are shown as points rather than trajectories because they remain noisy during early denoising steps.
Robust traffic simulators are crucial for developing and testing autonomous vehicles to reduce the costly, labor-intensive real-world data collection process and the need for physical presence on the road. However, existing simulators require agents' initial states to generate trajectories, which limits scalability and diversity due to restrictions on the given initial states. While data-driven agent initialization has been widely studied, the generated initial states are not interpretable in terms of why the agents are initialized at those specific locations. Given known initial states, trajectory generation is also a challenging problem, as the model must learn the variability of the destination and how agents should reach it over time. In this paper, we propose TrafficDiffuser, a top-down traffic scenario generation framework that generates high-level traffic scenarios, defined by initial and goal state pairs, by jointly modeling them. The high-level scenario generation makes initial states better interpretable and reduces trajectory generation into as simple as an infilling problem. We demonstrate how the generated high-level traffic scenarios can be used, including constraining based on different trajectory modes and integrating them with existing trajectory generation models. We conduct extensive experiments on the Argoverse 2 motion prediction dataset to evaluate how well the generated outputs capture real-world distributions. In addition to generating goal states, TrafficDiffuser outperforms the next-best approach for agent initialization, reducing speed distribution distance by 55.3% and the off-road rate by 2.8%.
Da Saem Lee, Yash Vardhan Pant, Sebastian Fischmeister
Department of Electrical and Computer Engineering, University of Waterloo, Waterloo, Canada
Building accurate decision-support tools for next-generation air traffic control requires robust trajectory prediction models. We present a flow-matching architecture trained exclusively on historical aircraft trajectories, with no route labels or chart supervision. Trained on 1.15 million Automatic Dependent Surveillance-Broadcast trajectory windows collected over the San Francisco Bay Area, the model generates aircraft trajectory distributions that closely match historical traffic, reproducing known airspace structure around San Francisco Airport such as the shape of SFO's published NIITE FOUR departure procedure. Our model is trained directly on the native, irregular ADS-B sampling interval. Trajectory prediction is cast as sequence inpainting using a block-causal Transformer that denoises future state tokens conditioned on the observed history using Conditional Flow Matching or Denoising Diffusion Probabilistic Models. We compare our architecture against constant-velocity, deterministic-Long Short Term Memory, and Conditional Variational Autoencoders baselines. At matched parameter count, CFM outperforms DDPM by 11-26% in minADE@20, and both generative objectives surpass the CVAE baseline by 31-41%. We further show that the error degrades gracefully with prediction horizon, and the architecture remains effective when retrained on temporally decimated feeds. Lastly, we sample K independent completions, yielding spatial probabilistic occupancy estimates that can serve as input to downstream conflict-risk estimation.
Mathurin Petit, Emir Torun, Louis Brusset +2
École Polytechnique, Palaiseau, 91128, France · Technische Universität Berlin, Berlin, 10623, Germany · Mines Paris–PSL University, Paris, 75006, France +2
Safety-critical traffic scenario generation is essential for evaluating autonomous driving systems under rare but high-risk interactions. Existing diffusion-based methods offer strong controllability in closed-loop generation, but their iterative denoising process is computationally expensive and may accumulate sampling and guidance errors over long rollouts, causing unrealistic motion artifacts such as jitter, abnormal acceleration, and off-road behavior. To address these issues, we propose RiskFlow, a closed-loop safety-critical multi-agent traffic generation framework that formulates future trajectory generation as transport in the action space. Instead of relying on iterative denoising, RiskFlow learns an average velocity field over a finite interval to transform Gaussian action sequences into future acceleration and yaw-rate commands with a single forward pass, using a JVP-based objective for efficient and stable training. At test time, RiskFlow applies output-space guidance to the generated actions, steering selected critical agents toward risky interactions while regularizing off-road behavior, and reconstructs physically feasible trajectories through vehicle dynamics. Experiments on nuScenes with tbsim closed-loop evaluation show that RiskFlow achieves a strong adversariality-realism trade-off across multi-agent and long-horizon settings. Compared with representative baselines, RiskFlow consistently improves realism while maintaining competitive safety-critical generation capability, and substantially reduces inference time for evaluation.
Qi Lan, Yining Tang, Yu Shen +4
College of Mechanical and Vehicle Engineering Chongqing University, Chongqing, China