Ellipse Loss for Scene-Compliant Motion Prediction
This work addresses the critical need for safe and efficient autonomous vehicle operations by improving trajectory prediction to obey map constraints, though it is incremental as it builds on an existing state-of-the-art model.
The paper tackles the problem of motion prediction for self-driving vehicles by introducing an ellipse loss that penalizes off-road predictions to ensure scene compliance, resulting in more accurate and realistic trajectory predictions as demonstrated on large-scale autonomous driving data.
Motion prediction is a critical part of self-driving technology, responsible for inferring future behavior of traffic actors in autonomous vehicle's surroundings. In order to ensure safe and efficient operations, prediction models need to output accurate trajectories that obey the map constraints. In this paper, we address this task and propose a novel ellipse loss that allows the models to better reason about scene compliance and predict more realistic trajectories. Ellipse loss penalizes off-road predictions directly in a supervised manner, by projecting the output trajectories into the top-down map frame using a differentiable trajectory rasterizer module. Moreover, it takes into account actor dimensions and orientation, providing more direct training signals to the model. We applied ellipse loss to a recently proposed state-of-the-art joint detection-prediction model to showcase its benefits. Evaluation on large-scale autonomous driving data strongly indicates that the method allows for more accurate and more realistic trajectory predictions.