Scene graph simplification improves robot task planning efficiency
An Information-Space Perspective to Scene Graph Sufficiency for Robotic Task Planning
Robotics
Summary
Robots need detailed maps of their surroundings, including objects and how they relate, to plan tasks effectively. But these maps, called scene graphs, can get very big and slow down planning. The authors explain how to simplify these maps without losing important information by merging or removing parts based on the specific task. They show when these simpler maps still let the robot plan successfully, and when they don’t. This helps make robot planning faster while keeping plans reliable.
What this means in practice
- •For robotics engineers: Create compact environment models that speed up planning while ensuring robot tasks remain feasible.
- •For autonomous vehicle developers: Implement task-focused environment representations to streamline decision-making in navigation and interaction.
A theory result. No direct application yet.
Authors
Başak Sakçak, Francesco Verdoja
Abstract
Planning in complex environments requires task specifications grounded in representations that capture objects, relations, and affordances; scene graphs meet this need, but their size in large environments hinders efficient planning. While task-aware pruning and hierarchical abstractions have been explored, a general, task-centric formalization of what constitutes a sufficient scene graph for planning remains open. This paper provides such a formalization by modeling planning over scene graphs within an information-spaces framework through the definition of scene graph transition systems and relevant action semantics for navigation and manipulation. We then introduce derived scene graphs via information mappings that merge and prune nodes and induce quotient transition systems augmented with motion primitives to capture higher-level actions over merged graph nodes. Sufficiency is characterized by two conditions: (i) the information mapping yields a deterministic quotient, and (ii) the task is well-posed over derived traces, ensuring plans found on the derived model are feasible on the maximal system. We illustrate the framework using a task over an example environment, showing both sufficient and insufficient reduced scene graphs.