Time to goal metric improves planning success for robotic navigation and manipulation
Time-to-Goal: Learned Temporal Metrics for Long-Horizon Planning with JEPA World Models
RoboticsComputer Vision and Pattern Recognition
Summary
Planning how to reach a goal point far in the future can be tricky if the robot’s internal sense of distance doesn’t reflect how close it is to the goal over time. The authors show that standard methods using simple distances in a hidden space don’t always tell time order clearly. They create a new way to learn a “time-to-goal” measurement that better ranks states according to how soon the goal will be reached. This learned metric helps the robot plan better and succeed more often on tasks involving navigation and manipulation.
What this means in practice
- •For roboticists: Use learned time-to-goal metrics to improve planning and control in robots navigating complex environments.
- •For autonomous vehicle teams: Enhance long-horizon route planning by incorporating temporal ordering into latent state representations for better decision-making.
Authors
Ruilong Yu, Fei Ye
Abstract
Joint-Embedding Predictive Architectures (JEPAs) support planning by comparing predicted latent states with an encoded goal. For this comparison to guide temporal progress, states closer in time to the goal should receive lower costs. However, training representations for prediction and distribution regularization does not directly enforce this ordering. We examine this mismatch in flat LeWorldModel (LeWM) by analyzing how its Euclidean goal cost changes with temporal separation. Our analysis identifies conditions under which states at different long offsets receive similar average costs, weakening their temporal distinction. To learn the missing ordering explicitly, we introduce Time-to-Goal (T2G), a lightweight metric trained on latent triples from offline trajectories. A margin ranking loss encourages smaller distances for temporally nearer states, requiring no rewards or policy labels. The learned metric then replaces the Euclidean planning cost while the JEPA encoder and predictor remain frozen. Diagnostics on AntMaze and FourRoom show that T2G preserves stronger temporal contrast and improves near--far ranking at long offsets. Across four navigation and manipulation benchmarks, T2G improves the average planning success rate from 32.7\% (flat LeWM) to 63.0\% and approaches or even surpasses the performance of the hierarchical H-JEPA baselines. These results show that learning temporal ordering for the goal cost can substantially improve latent planning even with a flat world model.