Courteous Anticipation: Improving Long-Lived Task Planning in Persistent Shared Environments

2026-07-22

AI Summary

In continuously shared environments, conventional task planners suffer from elevated cumulative costs over long task sequences due to their inability to anticipate future tasks and disregard constraints imposed by other robots. This work proposes a courteous, forward-looking planning approach that jointly optimizes current and expected future costs by integrating a model-driven planner—which generates candidate action sequences—with individual future cost estimators independently learned by each robot. To circumvent the combinatorial explosion associated with joint trajectory modeling, the method employs a factorized representation, enabling modular deployment wherein newly added robots require only training of their own dedicated cost estimator. Experimental results in domestic and restaurant scenarios demonstrate that the proposed method reduces total costs by 10.43%/4.03% and 17.41%/13.24%, respectively, compared to myopic and selfish forward-looking baselines.

Abstract

We consider a task planning scenario in which robots sharing a persistent environment are assigned tasks one at a time from a held-out sequence. Standard task planners, lacking foresight of future tasks and inconsiderate of others' constraints, solve each task in isolation, leaving terminal states that increase future cost for all, side effects that compound over lengthy task sequences. To reduce cost over the sequence, a robot must anticipate how its actions now may impact performance on future tasks for all robots sharing the environment. Therefore, we present courteous anticipatory planning, wherein a model-based planner proposes candidate plans and selects the one that jointly minimizes immediate cost and aggregated expected future cost across all robots, estimated via independent per-robot learned estimators. This factored formulation avoids combinatorial joint rollouts and supports modular deployment: adding a robot requires only training its own estimator. We evaluate in two persistent PDDL domains, a home environment with robots that have similar capabilities but different responsibilities, and a restaurant environment where robots' distinct capabilities create states that other robots lack the capability to resolve. During lengthy task sequences, our planner reduces total cost by 10.43% versus myopic and 4.03% versus selfish anticipatory planning in a two-robot home environment and by 17.41% and 13.24%, respectively, in a three-robot restaurant.

Problem

• task planning
• persistent environments
• multi-robot systems
• future cost anticipation
• shared environment

Innovation

• courteous anticipatory planning
• persistent shared environments
• multi-robot task planning
• learned cost estimators
• factored formulation