Abstract:Detection of slip during object grasping and manipulation plays a vital role in object handling. Existing solutions primarily rely on visual information to devise a strategy for grasping. However, for robotic systems to attain a level of proficiency comparable to humans, especially in consistently handling and manipulating unfamiliar objects, integrating artificial tactile sensing is increasingly essential. We introduce a novel physics-informed, data-driven approach to detect slip continuously in real time. We employ the GelSight Mini, an optical tactile sensor, attached to custom-designed grippers to gather tactile data. Our work leverages the inhomogeneity of tactile sensor readings during slip events to develop distinctive features and formulates slip detection as a classification problem. To evaluate our approach, we test multiple data-driven models on 10 common objects under different loading conditions, textures, and materials. Our results show that the best classification algorithm achieves a high average accuracy of 95.61%. We further illustrate the practical application of our research in dynamic robotic manipulation tasks, where our real-time slip detection and prevention algorithm is implemented.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper aims to solve the problem of detecting slip during the process of robots grasping and manipulating objects. Specifically, the paper proposes a new method based on tactile perception for real - time and continuous detection of whether an object slips when being grasped. #### Background and Problem Description 1. **Limitations of Existing Solutions**: - Most current solutions mainly rely on visual information to formulate grasping strategies. However, for robotic systems, to achieve a proficiency similar to that of humans, especially when dealing with and operating unfamiliar objects, visual information alone is not sufficient. - Integrating artificial tactile perception is becoming increasingly important to achieve more efficient object grasping and manipulation. 2. **Research Objectives**: - Develop a data - driven method based on physical principles to achieve real - time slip detection. - Use the GelSight Mini optical tactile sensor to obtain tactile data and extract features by analyzing the changes in the contact force field and its entropy. - Transform the slip - detection problem into a classification problem and evaluate the performance of different machine - learning models on multiple objects. 3. **Practical Applications**: - In dynamic robotic manipulation tasks, this method can detect and prevent slip in real - time, thereby improving the stability and reliability of the robotic system. #### Main Contributions - **Innovative Feature - Extraction Method**: Use the entropy change in tactile images as a key feature for slip detection without prior knowledge of the object or grasping conditions. - **High - Precision Detection Algorithm**: Verified by experiments, the best classification algorithm has achieved an average accuracy of 95.61%. - **Wide Applicability**: This method shows good generalization ability on objects of different shapes, materials, and textures and is suitable for various practical scenarios. Through these efforts, this research provides a more reliable and efficient means of slip detection for robotic systems, which helps to improve the operational performance of robots in complex environments.

Learning to Detect Slip through Tactile Estimation of the Contact Force Field and its Entropy

Learned Slip-Detection-Severity Framework using Tactile Deformation Field Feedback for Robotic Manipulation

Neuromorphic Event-Based Slip Detection and suppression in Robotic Grasping and Manipulation

Tactile-Based Slip Detection Towards Robot Grasping.

Maintaining Grasps within Slipping Bound by Monitoring Incipient Slip

Leveraging distributed contact force measurements for slip detection: a physics-based approach enabled by a data-driven tactile sensor

Slip Detection with Combined Tactile and Visual Information

Under Pressure: Learning to Detect Slip with Barometric Tactile Sensors

Learning to Detect Slip with Barometric Tactile Sensors and a Temporal Convolutional Neural Network

Robust Learning-Based Incipient Slip Detection using the PapillArray Optical Tactile Sensor for Improved Robotic Gripping

Incipient Slip Detection Method for Soft Objects with Vision-Based Tactile Sensor

Slip Detection in Prosthetic Hand Grasping by Using the Discrete Wavelet Transform Analysis

Finger‐Skin‐Inspired Flexible Optical Sensor for Force Sensing and Slip Detection in Robotic Grasping

A Robust Incipient Slip Detection Method with Vision-Based Tactile Sensor Based on Local Deformation Degree

Slip Detection: Analysis and Calibration of Univariate Tactile Signals

Flexible Tactile Sensor Array for Slippage and Grooved Surface Recognition in Sliding Movement

Learning to Predict Friction and Classify Contact States by Tactile Sensor

Incipient Slip Detection Method With Vision-Based Tactile Sensor Based on Distribution Force and Deformation

A model-free approach to fingertip slip and disturbance detection for grasp stability inference

Flexible Tactile Sensor Array for Distributed Tactile Sensing and Slip Detection in Robotic Hand Grasping

Perception, Control and Hardware for In-Hand Slip-Aware Object Manipulation with Parallel Grippers