Search papers, labs, and topics across Lattice.
This paper introduces an innovative mode-world weighted regression loss to enhance multi-agent trajectory prediction for automated vehicles, addressing the shortcomings of previous prediction-based and anchor-based methods. By mitigating mode collapse and improving world ranking and top-1 confidence, the approach significantly enhances the accuracy of motion predictions. Experimental results demonstrate that this method outperforms existing techniques, achieving first place in the Argoverse 2 multi-agent motion forecasting benchmark.
Mode-world weighted regression not only mitigates mode collapse but also boosts prediction accuracy, setting a new benchmark in multi-agent trajectory forecasting.
Multi-agent motion prediction is essential for automated vehicles to understand the intentions of surrounding vehicles. However, previous prediction-based and anchor-based methods have limitations in mode diversity and prediction accuracy, respectively. These limitations may cause inadequate safety assessments and behavioral deviations in automated vehicles. To address this issue, a mode-world weighted regression loss is proposed to bridge the gap between these features. Specifically, this approach mitigates mode collapse while simultaneously improving world ranking and top-1 confidence. Furthermore, the proposed iterative decoder improves prediction accuracy by recurrently and segmentally generating trajectories. Experimental results show the proposed method ranks first in the Argoverse 2 multi-agent motion forecasting benchmark against other methods.
