Uncertainty-Aware Trajectory Forecasting from Imperfect Tracking
2026-08-31 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors point out that most models predicting future paths of moving objects assume perfect input data, but real inputs from trackers are noisy and uncertain. They treat this uncertainty as useful information by representing each observed position as a fuzzy estimate with a Gaussian distribution that accounts for localization errors and data association issues. By doing this, their model not only predicts future positions as probabilities instead of exact points but also stays robust when the input data is noisy. They tested their approach on several datasets and found improvements in both accuracy and the quality of uncertainty estimates.
trajectory forecastingmulti-object trackinglocalization uncertaintydata associationGaussian distributionOrnstein-Uhlenbeck processknowledge distillationprobabilistic forecastingdisplacement accuracystructured noise
Authors
Stephane Da Silva Martins, Victor Petrovic, Emanuel Aldea, Sylvie Le Hégarat-Mascle
Abstract
Most trajectory forecasting models are trained on clean annotated histories, and are often evaluated under the same idealized assumption, although practical deployments rely on trajectories produced by imperfect multi-object trackers. The real-world observations exhibit localization jitter, missed or unstable detections, and data-association ambiguity, which are usually either ignored or removed through denoising. This paper instead treats tracking-derived reliability cues as an informative signal to be propagated to the predictor. We propose a plug-in uncertainty-aware formulation in which each observed state is encoded as an uncertain state representation, modeled by a Gaussian distribution whose covariance combines detection-level localization uncertainty and association-level ambiguity through the law of total variance. Existing backbones are adapted with minimal architectural changes: input trajectories are represented as Gaussian observations, and predicted trajectories are produced as Gaussian forecasts rather than deterministic coordinates. To train predictors that remain robust under structured observation noise, we combine temporally correlated Ornstein-Uhlenbeck perturbations with response-based knowledge distillation from a teacher trained on clean trajectories. Experiments on Oxford Town Centre and VIRAT using real tracker outputs, together with a complementary ETH/UCY pseudo-detection protocol, show that the proposed formulation improves displacement accuracy and the reliability-sharpness trade-off of probabilistic forecasts.