FutureLift improves motion forecasts by linking proposals in future graphs
FutureLift: Lifting Trajectory Proposals into Bidirectional Future Graphs for Motion Forecasting
Computer Vision and Pattern RecognitionMachine Learning
Summary
Predicting where vehicles will move next is important to keep autonomous cars safe, but current methods still make errors. The authors introduce FutureLift, a tool that takes initial guesses about vehicle movements and creates a connected graph of different possible futures. This graph can look both forward and backward in time to improve the predictions in one step. Tested on major datasets, FutureLift enhances existing forecasts and beats other refinement methods without needing extra data like LiDAR.
What this means in practice
- •For autonomous vehicle developers: Enhance existing vehicle motion forecasts by refining proposals using graph-based context for safer driving decisions.
- •For robotics software engineers: Integrate FutureLift to improve multimodal motion predictions when coordinating multiple agents in dynamic environments.
Authors
Daniel Grimm, J. Marius Zöllner
Abstract
Accurate motion prediction is critical for safe autonomous driving, yet even strong forecasters leave residual errors in their multimodal proposals. Existing refiners compress complete proposals or update their segments sequentially, preventing the full proposed future from serving as explicit relational context for correction. We introduce FutureLift, a backbone-agnostic refiner that represents each multimodal proposal as a heterogeneous future graph, with spatially grounded temporal nodes connected to agent histories, map elements, alternative modes, and, for joint prediction, other agents future trajectories. Bidirectional temporal edges allow later evidence to revise earlier segments in a single pass, and the updated nodes predict residual coordinate and confidence corrections. Requiring only proposals, scores, and scene context, FutureLift provides a common refinement interface for structurally different forecasting backbones. On Argoverse2 and the Waymo Open Motion Dataset, it consistently improves frozen proposals and outperforms native and specialized refiners across marginal and joint forecasting. On the WOMD motion-prediction leaderboard, FutureLift ranks first in minADE (0.5049m) and minFDE (0.9942m) among the compared standard-input methods, without LiDAR or ensembling, while achieving a competitive 0.4874 Soft mAP.