[Paper Review] Estimation and Exploitation of Objects' Inertial Parameters in Robotic Grasping and Manipulation: A Survey

Nikos Mavrakis, Rustam Stolkin|arXiv (Cornell University)|Nov 11, 2019

Robot Manipulation and LearningEngineering98 references54 citations

TL;DR

This survey proposes a comprehensive categorization of methods for estimating and exploiting object inertial parameters—mass, center of mass, and inertia tensor—in robotic grasping and manipulation. It classifies approaches into purely visual, exploratory, and fixed-object estimation methods, demonstrating how inertial knowledge enhances grasp planning, manipulation control, and dual-arm coordination, with key contributions in methodological synthesis and application mapping across robotics domains.

ABSTRACT

Inertial parameters characterise an object's motion under applied forces, and can provide strong priors for planning and control of robotic actions to manipulate the object. However, these parameters are not available a-priori in situations where a robot encounters new objects. In this paper, we describe and categorise the ways that a robot can identify an object's inertial parameters. We also discuss grasping and manipulation methods in which knowledge of inertial parameters is exploited in various ways. We begin with a discussion of literature which investigates how humans estimate the inertial parameters of objects, to provide background and motivation for this area of robotics research. We frame our discussion of the robotics literature in terms of three categories of estimation methods, according to the amount of interaction with the object: purely visual, exploratory, and fixed-object. Each category is analysed and discussed. To demonstrate the usefulness of inertial estimation research, we describe a number of grasping and manipulation applications that make use of the inertial parameters of objects. The aim of the paper is to thoroughly review and categorise existing work in an important, but under-explored, area of robotics research, present its background and applications, and suggest future directions. Note that this paper does not examine methods of identification of the robot's inertial parameters, but rather the identification of inertial parameters of other objects which the robot is tasked with manipulating.

Motivation & Objective

• To systematically categorize existing methods for estimating object inertial parameters in robotics.
• To analyze the strengths and limitations of purely visual, exploratory, and fixed-object estimation techniques.
• To demonstrate the critical role of inertial parameters in enhancing robotic grasping and manipulation performance.
• To identify gaps in current research and suggest future directions combining data-driven and model-based estimation.

Proposed method

• Classifies inertial parameter estimation into three categories based on robot-object interaction: purely visual (using geometry and prior knowledge), exploratory (active interaction via pushes or tilts), and fixed-object (firmly grasped or fixed for measurement).
• Reviews model-based estimation techniques that use kinematic and dynamic equations to infer inertial parameters from measured forces, torques, and motion.
• Examines data-driven approaches, including machine learning, for improving estimation accuracy, especially in ill-posed scenarios like single-pushing estimation.
• Analyzes sensing modalities such as depth sensors, force/torque sensors, and audio feedback, and their role in data acquisition for inertial parameter estimation.
• Maps estimation methods to specific robotic applications, including in-hand manipulation, bin-picking, and dual-arm coordination.
• Highlights the integration of inertial parameters into control frameworks, such as closed-loop dynamics models and torque-based planning.

Experimental results

Research questions

• RQ1 How do humans perceive and use inertial parameters, and what can robotics learn from this?
• RQ2 What are the three main categories of robotic inertial parameter estimation, and how do they differ in interaction level and environmental suitability?
• RQ3 How do inertial parameters improve performance in robotic grasping and manipulation tasks?
• RQ4 What are the key limitations of current estimation methods, especially in estimating full 3D inertia tensors?
• RQ5 How can future methods combine visual priors, exploratory actions, and data-driven learning to improve inertial estimation?

Key findings

• Purely visual methods can estimate all inertial parameters under strong assumptions of known density distribution, making them suitable for industrial environments with prior object knowledge.
• Exploratory methods enable autonomous inertial estimation in unknown environments but often fail to estimate full 3D inertia tensors without multi-axis motion, limiting accuracy.
• Fixed-object methods achieve high accuracy by using force/torque sensors during firm grasping or fixation, making them ideal for controlled industrial settings.
• Inertial parameters significantly improve grasp stability, re-grasping, and in-hand manipulation by enabling accurate dynamic modeling and torque prediction.
• Applications such as bin-picking, dual-arm manipulation, and push-based reorientation rely critically on accurate inertial parameters for effective task execution.
• Combining visual priors with exploratory or data-driven methods offers a promising path to solving the ill-posed problem of 3D inertial estimation from minimal interaction.