Goal Sets, Not Goal States: Queryable Robot Goals through Goal-Set Hindsight Relabeling
2026-06-08 • Robotics
Robotics
AI summaryⓘ
The authors found that a common trick called hindsight relabeling, which usually treats the robot's final state as the exact goal, can be too strict when only part of the state really matters for success. They introduced Goal-Set Hindsight Relabeling (GS-HER), which allows the robot to consider flexible sets of goals based on specific criteria, instead of fixed single goals. This approach improves robot learning by ignoring unnecessary details and lets one trained model handle different types of goals without needing retraining. Overall, their method makes offline robot learning more adaptable and efficient.
Hindsight RelabelingOffline Robot LearningGoal-Conditioned Reinforcement LearningPredicate LogicGoal SetsGoal RelabelingRobot LearningGeneralizationOffline Reinforcement LearningOGBench
Authors
Carlos Vélez García, Miguel Cazorla, Jorge Pomares
Abstract
Hindsight relabeling usually turns achieved future states into exact goals, which can overconstrain offline robot learning when task success depends only on a subset of the state. We propose Goal-Set Hindsight Relabeling (GS-HER), a predicate-level generalization of HER in which achieved states certify query-defined goal sets rather than singleton goal states. A binary query specifies which variables define success, making the goal predicate an inference-time input while leaving the underlying offline GCRL algorithm unchanged. Across OGBench tasks and five offline goal-conditioned learners, GS-HER improves performance when full-state goals are bottlenecked by nuisance dimensions and turns hindsight relabeling into a reusable goal interface: one checkpoint can answer multiple robot goal predicates without retraining.