Raising context awareness in motion forecasting
This work addresses a specific issue in trajectory prediction for autonomous driving, offering incremental improvements to existing methods.
The paper tackles the problem of motion forecasting models overly relying on agent dynamics instead of semantic context, introducing CAB to promote context use and new metrics for temporal consistency. It achieves evaluation on the nuScenes benchmark and a difficult subset, with code made available.
Learning-based trajectory prediction models have encountered great success, with the promise of leveraging contextual information in addition to motion history. Yet, we find that state-of-the-art forecasting methods tend to overly rely on the agent's current dynamics, failing to exploit the semantic contextual cues provided at its input. To alleviate this issue, we introduce CAB, a motion forecasting model equipped with a training procedure designed to promote the use of semantic contextual information. We also introduce two novel metrics - dispersion and convergence-to-range - to measure the temporal consistency of successive forecasts, which we found missing in standard metrics. Our method is evaluated on the widely adopted nuScenes Prediction benchmark as well as on a subset of the most difficult examples from this benchmark. The code is available at github.com/valeoai/CAB