Learning to Double Guess: An Active Perception Approach for Estimating the Center of Mass of Arbitrary Objects

2025-02-04

AI Summary

Estimating the center of mass (CoM) of arbitrary objects in unstructured environments remains challenging due to sensor noise and modeling inaccuracies inherent in conventional single-interaction approaches. To address this, we propose U-GRAPH, an uncertainty-guided rotational active perception framework. U-GRAPH integrates a Bayesian neural network for predictive uncertainty quantification, an action-scoring network, and a tactile-motor closed-loop controller to dynamically optimize multi-step, information-gain-driven interactions. Crucially, it introduces a grid-search-assisted, uncertainty-driven decision mechanism that enables high-accuracy and robust CoM estimation for unseen, complex real-world objects—even when trained on small-scale, low-diversity datasets. Experiments demonstrate that U-GRAPH significantly outperforms existing methods in generalization capability and cross-object transfer performance.

Technology Category

• Reasoning under Uncertainty: Graphical Models
• Intelligent Robots: State Estimation
• Multiagent Systems: Multiagent Systems under Uncertainty

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Manipulating arbitrary objects in unstructured environments is a significant challenge in robotics, primarily due to difficulties in determining an object's center of mass. This paper introduces U-GRAPH: Uncertainty-Guided Rotational Active Perception with Haptics, a novel framework to enhance the center of mass estimation using active perception. Traditional methods often rely on single interaction and are limited by the inherent inaccuracies of Force-Torque (F/T) sensors. Our approach circumvents these limitations by integrating a Bayesian Neural Network (BNN) to quantify uncertainty and guide the robotic system through multiple, information-rich interactions via grid search and a neural network that scores each action. We demonstrate the remarkable generalizability and transferability of our method with training on a small dataset with limited variation yet still perform well on unseen complex real-world objects.

Problem

• Estimating arbitrary objects' center of mass
• Overcoming Force-Torque sensor inaccuracies
• Enhancing robotic manipulation with active perception

Innovation

• Bayesian Neural Network
• Active perception
• Uncertainty quantification