Spatial Language Likelihood Grounding Network for Bayesian Fusion of Human-Robot Observations

TL;DR

This paper introduces FP-LGN, a neural network that grounds spatial language in map features to enable uncertainty-aware fusion of human and robot observations, improving collaborative task performance.

Contribution

The paper presents a novel Feature Pyramid Likelihood Grounding Network that learns spatial language grounding with uncertainty estimation for robotic perception.

Findings

FP-LGN matches expert rules in Negative Log-Likelihood

The model shows greater robustness with lower uncertainty

Fusion of human language and robot sensors improves task performance

Abstract

Fusing information from human observations can help robots overcome sensing limitations in collaborative tasks. However, an uncertainty-aware fusion framework requires a grounded likelihood representing the uncertainty of human inputs. This paper presents a Feature Pyramid Likelihood Grounding Network (FP-LGN) that grounds spatial language by learning relevant map image features and their relationships with spatial relation semantics. The model is trained as a probability estimator to capture aleatoric uncertainty in human language using three-stage curriculum learning. Results showed that FP-LGN matched expert-designed rules in mean Negative Log-Likelihood (NLL) and demonstrated greater robustness with lower standard deviation. Collaborative sensing results demonstrated that the grounded likelihood successfully enabled uncertainty-aware fusion of heterogeneous human language observations and robot sensor measurements, achieving significant improvements in human-robot collaborative task performance.