Dexterous Planar Pushing under Uncertain Object Properties: A Contact-Aware Goal-Oriented Approach
Yongseok Lee, Keehoon Kim
AI summary
Problem
Robotic pushing struggles with hybrid contact dynamics and uncertain object properties, often relying on restrictive non-separation assumptions or fixed trajectories that limit dexterity and generalization.
Approach
The framework introduces a contact-aware generalized velocity-motion model that captures sticking, sliding, and separation modes, integrated with MPPI control and UKF-based online estimation to adapt to real-world uncertainties.
Key results
- Validated C-GVMM accuracy on the MIT pushing dataset across diverse contact modes
- Achieved high success rates in re-positioning, re-orienting, and obstacle avoidance under unknown object properties
- Demonstrated robust real-world pushing on a 7-DoF manipulator without predefined trajectories
- Outperformed fixed-parameter MPPI and offline planners in trajectory optimality and success rate under uncertainty
Why it matters
Enables reliable, dexterous non-prehensile manipulation for real-world robots handling ungraspable or poorly characterized objects.
Abstract
Robotic pushing is a versatile non-prehensile ma- nipulation skill that enables robots to handle ungraspable ob- jects without specialized tools. This paper introduces a contact- aware, goal-oriented pushing framework that achieves dexter- ous and robust manipulation by explicitly allowing free-motion of the end-effector. Central to our approach is the contact-aware generalized velocity–motion model (C-GVMM), which captures the relationship between pusher velocity and slider motion across all contact modes, including separation. Unlike prior methods that rely on predefined trajectories or fixed contact- mode sequences, our framework enables seamless transitions among sticking, sliding, and separating modes. Building upon C-GVMM, we employ Model Predictive Path Integral (MPPI) control to generate goal-directed actions, and UKF-based online estimation to handle the uncertain object properties in real- world setting. We validate our approach through both nu- merical simulations and real-robot experiments, demonstrating that the framework accomplishes diverse pushing tasks with more optimal pusher and slider motion with high success rates. These results demonstrate the practical viability of the proposed approach for real-world robotic pushing tasks.