Combined task and motion planning through an extensible planner-independent interface layer

Abstract

Abstract-The need for combined task and motion planning in robotics is well understood. Solutions to this problem have typically relied on special purpose, integrated implementations of task planning and motion planning algorithms. We propose a new approach that uses off-the-shelf task planners and motion planners and makes no assumptions about their implementation. Doing so enables our approach to directly build on, and benefit from, the vast literature and latest advances in task planning and motion planning. It uses a novel representational abstraction and requires only that failures in computing a motion plan for a high-level action be identifiable and expressible in the form of logical predicates at the task level. We evaluate the approach and illustrate its robustness through a number of experiments using a state-of-the-art robotics simulator and a PR2 robot. These experiments show the system accomplishing a diverse set of challenging tasks such as taking advantage of a tray when laying out a table for dinner and picking objects from cluttered environments where other objects need to be re-arranged before the target object can be reached.

Incremental Task and Motion Planning: A Constraint-Based Approach

“…Previous work applied a variety of notations to the modeling of discrete robot tasks: temporal logics [3,36,47], structured and natural language [46,59], logical knowledge bases [25,75], the Planning Domain Definition Language (PDDL) [30,61,72], and context-free grammars [14,58,79]. Each notation provides advantages for certain domains or proprieties, e.g., safety properties are easily specified with temporal logics, action effects with PDDL, and hierarchies with grammars.…”

“…Such incremental constraint updates also directly apply to additional geometric information from the motion layer. For example, the specific objects preventing an action could be estimated using MCR [35] or the heuristic method of [72], further improving generalization of (3).…”

“…The typical algorithms for independent task planning [37,39,43,67] and motion planning [44,53] are fundamentally different. Consequently, most TMP methods [9,36,69,72] perform task planning and motion planning as separate, possibly interleaved, phases. We isolate and discuss the specific requirements for task planning, motion planning, and their interface in order to perform efficient and robust TMP.…”

“…Task domains are typically defined in terms of states and actions using a variety of notations [25,36,72]. To support various notations, we define the notation-independent task domain as the following formal language:…”

“…Several related methods [17,18,19,30] extend Hierarchical Task Networks (HTN) [26] with geometric primitives, using shared literals to control backtracking between the task and motion layer. [72] interfaces off-the-shelf task planners with an optimization-based motion planner [70] using a heuristic to remove potentially-interfering objects. [57] formulates the motion side of TMP as a constraint satisfaction problem over a discretized, preprocessed subset of the configuration space.…”