FloAff-Kitchen: Bridging Navigation and Manipulation via Canonical and Progressive Floor Affordance Learning
2026-07-27 • Robotics
Robotics
AI summaryⓘ
The authors focus on helping robots figure out the best spots on the floor to stand so they can perform tasks successfully, beyond just being able to move around. They created a new way for robots to understand these spots by removing confusing details and learning from basic tasks to improve in more complex ones. They also built a test set with lots of different kitchen scenes to see how well their method works. Their experiments show their approach is better than older methods, and they checked that each part of their solution helps.
Floor AffordanceMobile ManipulationCanonical Representation LearningEgocentric PerceptionSpatial ReasoningMulti-modal PerceptionAffordance Prior LearningBenchmark DatasetTransfer LearningRobotics
Authors
Ping Zhong, Manling Teng, Tao Wu, Bolei Chen, Jiazhi Xia, Jianxin Wang
Abstract
Mobile manipulation requires robots to identify Floor Affordance (FloAff) that maximizes downstream manipulation success rather than merely ensuring navigation feasibility. FloAff prediction is a target-conditioned local spatial reasoning problem, yet existing methods suffer from representation ambiguity caused by irrelevant spatial context and arbitrary object orientations, while entangling shared and task-specific knowledge across heterogeneous manipulation skills. To address these challenges, we propose a unified framework for FloAff prediction from egocentric multimodal perception, consisting of canonical representation learning and progressive affordance prior learning. Specifically, we introduce a Canonical Floor Affordance Representation (CFAR), which learns canonical interaction geometry by preserving affordance-relevant local structure while eliminating nuisance spatial variations unrelated to robot base placement. We further propose Progressive Floor Affordance Learning (PFAL), which learns transferable FloAff priors from a foundation manipulation task and progressively adapts them to heterogeneous downstream manipulation skills. To facilitate systematic evaluation, we establish the first cross-scene, multi-view FloAff-Kitchen benchmark covering diverse manipulation skills, scene layouts, furniture styles, and viewpoints. Extensive experiments on three benchmark settings demonstrate that our method consistently outperforms strong baselines, while ablation studies validate the contribution of each proposed component. Project page: https://csu-hero-lab.github.io/FloAff-Kitchen_Web/