Likely, Light, and Accurate Context-Free Clusters-based Trajectory Prediction
This work addresses trajectory prediction for autonomous vehicles, presenting an incremental improvement with specific gains in accuracy and efficiency.
The paper tackles trajectory forecasting for autonomous road systems by proposing a multi-stage probabilistic approach with a new deep feature clustering method and distance-based ranking, achieving superior performance over context-free deep generative models and matching point estimators on the most probable trajectory.
Autonomous systems in the road transportation network require intelligent mechanisms that cope with uncertainty to foresee the future. In this paper, we propose a multi-stage probabilistic approach for trajectory forecasting: trajectory transformation to displacement space, clustering of displacement time series, trajectory proposals, and ranking proposals. We introduce a new deep feature clustering method, underlying self-conditioned GAN, which copes better with distribution shifts than traditional methods. Additionally, we propose novel distance-based ranking proposals to assign probabilities to the generated trajectories that are more efficient yet accurate than an auxiliary neural network. The overall system surpasses context-free deep generative models in human and road agents trajectory data while performing similarly to point estimators when comparing the most probable trajectory.