Behavior Composition as Fully Observable Non-Deterministic Planning

Abstract

The behavior composition problem involves the automatic synthesis of a controller able to “realize” (i.e., implement) a target behavior module by suitably coordinating a collection of partially controllable available behaviors. In this paper, we show that the existence of a composition solution amounts to finding a strong cyclic plan for a special non-deterministic planning problem, thus establishing the formal link between the two synthesis tasks. Importantly, our results support the use of non-deterministic planing systemsfor solving composition problems in an off-the-shelf manner. We then empirically evaluate three state-of-the-art synthesis systems (a domain-independent automated planner and two game solvers based on model checking techniques) on various non-trivial composition instances. Our experiments show that while behavior composition is EXPTIME-complete, the current technology is already able to handle instances of significant complexity. Our work is, as far as we know, the first serious experimental work on behavior composition.

Building Virtual Behaviors from Partially Controllable Available Behaviors in Nondeterministic Environments

“…Indeed, the composition problem requires an advanced conditional plan (with loops) that always guarantees all possible target requests to be "served," which is, ultimately, a (temporal) invariant property. What is more, as later proved by (Ramirez, Yadav, and Sardina 2013), such plans amount to strong-cyclic policies, the general solution concepts for FOND planning. Even more, the solutions obtained via the simulation technique developed in this work are akin to the so-called universal plans (Schoppers 1987), that is, plans representing every possible solution.…”

Non-Deterministic Planning With Conditional Effects

“…The newly introduced planner represents an important increase in the scope of problems that can be solved by non-deterministic planning techniques. This development is significant not only for FOND planning but for emerging applications that can exploit FOND planning in their computational core (e.g., (Ramírez, Yadav, and Sardiña 2013;Patrizi, Lipovetzky, and Geffner 2013)).…”

“…Fully observable non-deterministic (FOND) planning problems require the planner to consider every possible action outcome, often guaranteeing achievement of the goal under an assumption of fairness. Recent advances have improved the efficiency of FOND planning (Kuter et al 2008;Mattmüller et al 2010;Fu et al 2011;Muise, McIlraith, and Beck 2012), opening the door to new FOND-based techniques for a variety of applications, including behaviour composition (Ramírez, Yadav, and Sardiña 2013) and the synthesis of controllers satisfying temporally extended specifications (Patrizi, Lipovetzky, and Geffner 2013).…”

A Declarative Approach to Compact Controllers for FOND Planning via Answer Set Programming

“…In fully-observable non-deterministic (FOND) planning [8,6], actions with alternative possible effects are allowed at the representational level: the agent does not know which outcome will occur when it executes an action, but it will be able to observe the outcome once executed. Despite its conceptual simplicity, FOND planning has proven to be a very powerful framework, and it has been connected to very expressive problems in AI and CS, like generalized planning [31], service/behavior composition [25], and even general reactive synthesis [5].…”