PDDLStream: Integrating Symbolic Planners and Blackbox Samplers via Optimistic Adaptive Planning

By Caelan Reed Garrett, Tomas Lozano-Perez, Leslie Pack Kaelbling. International Conference on Automated Planning and Scheduling, 2018

353 citations recorded by the source corpus.

Many planning applications involve complex relationships defined on high-dimensional, continuous variables. For example, robotic manipulation requires planning with kinematic, collision, visibility, and motion constraints involving robot configurations, object poses, and robot trajectories. These constraints typically require specialized procedures to sample satisfying values. We extend PDDL to support a generic, declarative specification for these procedures that treats their implementation as black boxes. We provide domain-independent algorithms that reduce PDDLStream problems to a sequence of finite PDDL problems. We also introduce an algorithm that dynamically balances exploring new candidate plans and exploiting existing ones. This enables the algorithm to greedily search the space of parameter bindings to more quickly solve tightly-constrained problems as well as locally optimize to produce low-cost solutions. We evaluate our algorithms on three simulated robotic planning domains as well as several real-world robotic tasks.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/394e3995d2faf542104cb7e6ea43cec204253273

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

ManipulaTHOR: A Framework for Visual Object Manipulation (2021, background): “Past works for visual object manipulation tend to use modular designs to solve this problem [25, 30, 14] – for instance, employing object detection models, instance segmentation models, point cloud estimators, etc. and then feeding these outputs into a planner.”
ProgPrompt: Generating Situated Robot Task Plans using Large Language Models (2022, background): “For high-level planning, most works in robotics use search in a pre-defined domain [5], [6], [7].”
Hierarchical Planning for Long-Horizon Manipulation with Geometric and Symbolic Scene Graphs (2020, methodology): “Problem Size PDDLStream [10] Ours Task Motion Task Motion”
Text2Motion: from natural language instructions to feasible plans (2023, methodology): “To amortize planning costs, works learn sampling distributions [8, 38– 41], visual feasibility heuristics [42–44], low-level controllers [45, 46], or state sparsifiers [47, 48], from datasets of solutions computed by classical TAMP solvers.”
Open-vocabulary Queryable Scene Representations for Real World Planning (2022, background): “In task and motion planning, scene representations are often composed of predicates compatible with symbolic planners [32], [33].”
Online Replanning in Belief Space for Partially Observable Task and Motion Problems (2019, methodology): “We use the Focused algorithm [11] to solve each determinized PDDLStream problem.”
Integrated Task and Motion Planning (2020, methodology): “The focused algorithm presented in (72, 55, 73) plans using a mixed set of sampled values and free parameters (optimistic values), which represent values that are not yet available, but that might potentially be generated by a sampler.”
Planning with Diffusion for Flexible Behavior Synthesis (2022, methodology): “We train all methods on 10000 trajectories from demonstrations generated by PDDLStream (Garrett et al., 2020); rewards are equal to one upon successful stack placements and zero otherwise.”
Large Language Models as Commonsense Knowledge for Large-Scale Task Planning (2023, background): “In the early stage, many studies [1, 15, 14, 12] focused on task planning in a discrete state space with deterministic transitions.”