Temporal Planning with Intermediate Conditions and Effects

Abstract

Automated temporal planning is the technology of choice when controlling systems that can execute more actions in parallel and when temporal constraints, such as deadlines, are needed in the model. One limitation of several action-based planning systems is that actions are modeled as intervals having conditions and effects only at the extremes and as invariants, but no conditions nor effects can be specified at arbitrary points or sub-intervals.In this paper, we address this limitation by providing an effective heuristic-search technique for temporal planning, allowing the definition of actions with conditions and effects at any arbitrary time within the action duration. We experimentally demonstrate that our approach is far better than standard encodings in PDDL 2.1 and is competitive with other approaches that can (directly or indirectly) represent intermediate action conditions or effects.

Deciding Unsolvability in Temporal Planning under Action Non-Self-Overlapping

“…Intuitively, a plan execution can be seen as a sequence of events separated by time elapses. This is the same idea of "snap-actions" of PDDL 2.1 (Coles et al 2008) generalized to the ANML case where we have events that might not coincide to the starting or ending of actions (called Intermediate Conditions and Effects (ICE) (Valentini, Micheli, and Cimatti 2020)). Complexity and self-overlapping.…”

“…We start by formalizing the TP problem; we tackle TP problems admitting Intermediate Conditions and Effects (ICE) (Valentini, Micheli, and Cimatti 2020) and we use the same fragment of the ANML (Smith, Frank, and Cushing 2008) modeling language used in (Valentini, Micheli, and Cimatti 2020) with two additional assumptions needed for the problem to be decidable. We forbid infinite-domain fluents and we work under the ANSO assumption: we disallow any plan that exhibits a time instant when two instances of the same ground action are running.…”

“…We now present the TAMER-CTP planning algorithm, a complete heuristic-search based temporal planner that works by maintaining the temporal information symbolic with Difference Bound Matrices (DBMs) (Bengtsson and Yi 2004), while using explicit propositional states. TAMER-CTP inherits the subdivision of actions in time-points, the search procedure and the heuristic from TAMER-FTP (the planner presented by Valentini, Micheli, and Cimatti 2020), but it differs completely in the way it handles temporal constraints. While TAMER-FTP relies on Simple Temporal Networks (STNs) (Dechter, Meiri, and Pearl 1991) to add and check constraints between time-points, TAMER-CTP is based on DBMs, the data structure used in the context of TA to represent zones (Bengtsson et al 1995).…”

“…We propose two novel terminating approaches: the first approach is a reduction to model-checking, which encodes the problem in a symbolic timed transition system (Cimatti et al 2019) and uses the NUXMV (Cavada et al 2014) model-checker; the second approach is a dedicated Forward-Chaining Temporal Planner called TAMER-CTP, which combines a heuristic-search schema inspired by (Valentini, Micheli, and Cimatti 2020) with a symbolic representation of time using Difference Bound Matrices (DBM).…”

FAPE: a Constraint-based Planner for Generative and Hierarchical Temporal Planning

“…To the best of our knowledge no planner exists that support all ANML features and in particular, none that supports any of its hierarchical features. TAMER [VMC20] is a forward-chaining planner that supports a form of intermediate conditions and effects of the ANML language. More specifically, it supports conditions and effects constrained to be a fixed time-amount after the start or before the end of an action.…”