Summary
Planning systems usually need to be trained specifically for each task domain, meaning they can't easily work on new problems they haven't seen before. The authors created a new method called GADAR that can understand and act in new domains without extra training by recognizing the roles and relationships of actions rather than their names. This method works by embedding the domain's structure in three layers that help rank actions toward achieving goals. Tests show one GADAR model can handle many different domains almost as well as specialized models. While it still works worse than models trained specifically on each domain, GADAR does much better than random choices on new problems without needing retraining.
What this means in practice
- •For robotics engineers: Build robots that can plan actions in new environments without retraining by using a single policy that generalizes across tasks.
- •For game developers: Design game AI that adapts to different game rules and tasks without training models for each specific game scenario.
Abstract
Learned policies for classical planning act without search, but each is tied to the domain it was trained on: network modules per action schema, parameters per predicate, or vocabulary-sized encodings mean a trained policy cannot even be executed on a new domain's predicates and action schemas. We present GADAR (Generalizing Across Domains via Action Ranking), a policy that takes the domain description as input, π(s, G, D), and constructs grounded actions in domains never seen in training, without search. Symbols are described by the roles they play in the domain rather than by their names, in three layers that each extend the previous one: (i) a lifted domain layer embeds D in every state graph through structural signatures; (ii) a binding layer ties each ground action's applicability to the lifted preconditions it instantiates; and (iii) a chain layer states which schema produces what another needs and how far each schema stands from the goal. The same network therefore runs on every domain, seen or unseen. We evaluate zero-shot generalization leave-one-domain-out over eight benchmark domains. A single GADAR policy trained on seven domains solves 58.3% of their test instances, one point below the 59.3% of dedicated single-domain GADAR models, so one policy serves many domains at no cost in coverage; a multi-task control that keeps symbol identity in its weights, while strong on some of the domains it trains on, transfers to none of the held-out ones. Zero-shot, GADAR executes on every held-out domain and exceeds random choice under the same executor on the large test instances of all eight: it solves 22.6% of them on average, against 3.3% for random choice and 0.6% for the same representation without the chain layer, and reaches 68% on held-out Visitall. Zero-shot coverage nonetheless remains well below in-domain coverage on most domains, and we analyze where it is weakest.