Anticipatory Planning for Performant Long-Lived Robot in Large-Scale Home-Like Environments

TL;DR

This research introduces a model-based anticipatory task planning framework designed to scale to large-scale realistic environments and uses a graph neural network in particular via a representation inspired by a 3D scene graph to learn the essential properties of the environment crucial to estimating the state's expected cost.

Abstract

We consider the setting where a robot must complete a sequence of tasks in a persistent large-scale environment, given one at a time. Existing task planners often operate myopically, focusing solely on immediate goals without considering the impact of current actions on future tasks. Anticipatory planning, which reduces the joint objective of the immediate planning cost of the current task and the expected cost associated with future subsequent tasks, offers an approach for improving long-lived task planning. However, applying anticipatory planning in large-scale environments presents significant challenges due to the sheer number of assets involved, which strains the scalability of learning and planning. In this research, we introduce a model-based anticipatory task planning framework designed to scale to large-scale realistic environments. Our framework uses a GNN in particular via a representation inspired by a 3D Scene Graph to learn the essential properties of the environment crucial to estimating the state's expected cost and a sampling-based procedure for practical large-scale anticipatory planning. Our experimental results show that our planner reduces the cost of task sequence by 5.38% in home and 31.5% in restaurant settings. If given time to prepare in advance using our model reduces task sequence costs by 40.6% and 42.5%, respectively.

Figures (7)

• Figure 2: Home-Scale Anticipatory Planning: The service robot is tasked with serving a cup of water on a desk in a large home. Next, it might be asked to 'clean the jar' or 'wipe the table' from the task distribution. Myopic Approach: Completes the task with a lower immediate cost by using a cup to fetch and serve water. Our Approach: The robot cleans and fills up the jar with water and completes the task which is more costly now but reduces the cost of the future tasks thus reduces the overall cost of completing all the tasks.
• Figure 3: Schematic of our approach: (Left) Shows an overview of our approach described in \ref{['sec:search']} and (Right) shows the maps and representations for learning described in \ref{['sec:learning-3dsg']}
• Figure 4: Sampling procedure: We sample tasks for planning by augmenting the given task with other predicates only involving entities within a bounded region.
• Figure 5: Restaurant Domain Example with two tasks: Anticipatory Planning (left) reduces the total planning cost over Myopic Planning (middle) by anticipating the future tasks. Preparation in advance (right) further reduces the planning cost.
• Figure 6: Preparing the environments using our estimator reduces the average-cost per each task of the task-sequence compared to planning without preparation.