Hybrid Task Planning Grounded in Belief: Constructing Physical Copies of Simple Structures

Abstract

Symbolic planning methods have proved to be challenging in robotics due to partial observability and noise as well as unavoidable exceptions to rules that symbol semantics depend on. Often the symbols that a robot considers to support for planning are brittle, making them unsuited for even relatively short term use. Maturing probabilistic methods in robotics, however, are providing a sound basis for symbol grounding that supports using probabilistic distributions over symbolic entities as the basis for planning. In this paper, we describe a belief-space planner that stabilizes the semantics of feedback from the environment by actively interacting with a scene. When distributions over higher-level abstractions stabilize, powerful symbolic planning techniques can provide reliable guidance for problem solving. We present such an approach in a hybrid planning scheme that actively controls uncertainty and yields robust state estimation with bounds on uncertainty that can make effective use of powerful symbolic planning techniques. We illustrate the idea in a hybrid planner for autonomous construction tasks with a real robot system.

Autonomous Building of Structures in Unstructured Environments via AI Planning

“…Toussaint et al (2015) and Leidner et al (2013) all demonstrate PDDL being used to perform spatial reasoning in performing robotic tasks, but the spatial reasoning is abstract and of features such as objects 'on top' of other objects. Takahashi et al (2017) also demonstrates spatial reasoning in order to produce structures from cubic blocks. In their work, the spatial reasoning is performed abstractly.…”

Planning Robust Manual Tasks in Hierarchical Belief Spaces

“…Recent work has started to utilize ARcubes to form simple assemblies such as towers. In that work, a symbolic planner manipulated symbols grounded in belief by a belief space planner to resolve action pre-conditions and resource constraints in unstructured environments (Takahashi, Lanighan, and Grupen 2017). This system demonstrated how symbols grounded in belief lead to more reliable solutions than using maximum likelihood assumptions alone.…”

“…In each experiment, the raw materials (ARcubes) needed to construct the assembly are present, although their poses and identities are initially unknown to the robot. The first experiment compared the overall error at the conclusion of a tower assembly using the proposed method to a baseline system used in (Takahashi, Lanighan, and Grupen 2017). This baseline system lifts objects in the environment to symbolic representations and then executes a plan prescribed by a symbolic planner.…”

“…The features were required to be located in specific positions relative to a fixed frame defined by 3 environmental tags ("A", "B", and "C") as seen in Figure 3. We compared the proposed method with a baseline method (Takahashi, Lanighan, and Grupen 2017). When using the baseline, the system would fail when action outcomes did not match the expectations of the planner.…”