LENS: LLM-guided Environment Simplification for Planning and Control in Clutter

2026-07-21

AI Summary

Real-world, highly cluttered multi-object scenes pose significant challenges to existing robotic manipulation approaches due to dense object interactions, physical uncertainties, and task ambiguity. This work proposes the LENS framework, which introduces—for the first time—an LLM-guided adaptive scene abstraction mechanism that dynamically constructs task-oriented, simplified environment representations without human intervention. By iteratively merging or pruning scene entities in a closed-loop manner, LENS jointly optimizes classical planning, model-based control, and vision-language-action models. Experimental results demonstrate that LENS substantially enhances overall system performance across a variety of complex manipulation tasks.

Abstract

Despite recent advances in general-purpose robotic manipulation, real-world multi-object clutter remains challenging to handle for today's prevalent approaches. The problem scales in complexity due to more objects and collisions, more unpredictable contact physics, distractors, and task ambiguity. Bridging this gap to real-world deployment requires effective scene abstractions; yet today, producing such abstractions requires extensive task-specific manual engineering, which does not scale. These abstractions are costly to generate and difficult to adjust or fine-tune. We instead propose a plug-and-play fix to automatically generate scene-specific, task-specific, adaptively updating abstractions on top of existing planning and control stacks. LLM-guided Environment Simplification (LENS) produces a de-cluttered abstracted scene representation by merging (e.g., stacked objects) or pruning (e.g., distant objects) scene entities in a closed loop in response to task progress. These dynamic, task-relevant abstractions are versatile and easy to use. In our experiments, we show that LENS improves classical planning, model-based control, and a vision-language-action model, across a diverse set of highly cluttered manipulation scenes. Project website: https://lens-2026.github.io/.

Problem

• clutter
• scene abstraction
• robotic manipulation
• task ambiguity
• real-world deployment

Innovation

• LLM-guided abstraction
• environment simplification
• task-specific scene representation
• cluttered manipulation
• adaptive planning