Unifying Deep Predicate Invention with Foundation Models for Long Horizon Planning
Main Demo. UniPred performs long horizon planning with learned neural-symbolic world model.
Demo for other tasks...
Demo showing failure recovery.
UniPred continuously replans based on observation feedback, detects action failures during execution, and performs recovery replanning. Through a total of 13 steps, the system successfully completes the task.
Other Demos showing failure recovery...
Long horizon manipulation tasks are inherently challenging.Many unpredictable factors that can lead to failure, including planning failures and accumulated execution errors at each action step.
If you use UniPred in your research please cite the corresponding paper
TODO
• add paper citation
• add project webpage
TODO
• add license information
