Robot Grasp Planning for Safe Human Object Handovers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current robotic systems face challenges in reliably receiving objects from human operators without interfering with the human hand, leading to potential drops and injuries, especially in collaborative tasks where precise object transfer is required.
Innovation Solution
A vision-based system using depth cameras to generate point clouds, classify human hand poses, and plan robotic grasps that avoid contact with the human hand, enabling reliable and adaptive object transfer by segmenting the hand and object in the point cloud and selecting appropriate robotic grasps based on human hand poses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the robot selects a grasp that can reliably grip an object, then the gripping reliability is improved, but the human hand may be interfered with
Solution Approach 1:
The system segments the point cloud into hand points and object points using a trained neural network classifier. This segmentation allows the robot to identify the boundary between the human hand and the object, enabling it to select grasp poses that contact only the object and not the hand, thereby resolving the contradiction between reliable gripping and avoiding hand interference.
Solution Approach 2:
The patent introduces an intermediary computational process involving depth camera data, point cloud generation, and neural network-based classification to mediate between the robot's gripping action and the human hand. This intermediary system analyzes the spatial relationship between hand and object to determine safe grasp poses, preventing direct interference while maintaining gripping reliability.
2Device complexity
If the robot uses a simple grasp selection method, then the system complexity is reduced, but the handover success rate decreases
Solution Approach 1:
The patent replaces traditional mechanical or rule-based grasp selection methods with a neural network-based classification system. The neural network processes depth image data and point cloud information to automatically classify hand poses and generate appropriate grasp poses, substituting complex mechanical decision-making with intelligent algorithms that improve handover success while managing system complexity through software-based solutions.
3Productivity
If the robot approaches the object quickly, then the handover efficiency is improved, but the safety margin is reduced
Solution Approach 1:
The system performs preliminary classification of the hand pose and generation of appropriate grasp poses before the robot executes the handover action. By pre-processing the depth image data and determining the optimal grasp configuration in advance, the robot can then execute the handover quickly and safely, as the safe grasp pose has already been calculated and validated before movement begins.
Data Source
AI summary
A robotic control system directs a robot to take an object from a human grasp by obtaining an image of a human hand holding an object, estimating the pose of the human hand and the object, and determining a grasp pose for the robot that will not interfere with the human hand. In at least one example, a depth camera is used to obtain a point cloud of the human hand holding the object. The point cloud is provided to a deep network that is trained to generate a grasp pose for a robotic gripper that can take the object from the human's hand without pinching or touching the human's fingers.