Robotic Grasp Generation Using Virtual Clearance and Collision Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robotic grasp generation techniques are computationally expensive, generate low-quality grasps, and unnecessarily rule out potential grasps due to predicted collisions, which are often compensated for in real-world scenarios.
Innovation Solution
A method that optimizes grasp candidates using a virtual gripper surface with a clearance margin, followed by a signed distance field calculation to filter out collisions, resulting in a high-quality, collision-free grasp database for robotic pick-and-place operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing grasp generation techniques are used, then a database of grasps can be generated, but the computation is extremely expensive and slow
Solution Approach 1:
The grasp generation process is segmented into two distinct phases: (1) generating a large set of candidate grasps using simplified models, and (2) filtering and optimizing these candidates using more accurate collision detection. This segmentation allows the system to efficiently explore the grasp space while maintaining computational feasibility for quality assessment.
Solution Approach 2:
The method performs preliminary grasp generation using simplified models before applying computationally intensive collision detection. By pre-filtering candidates using faster approximation methods, the system reduces the number of grasps that require expensive detailed collision checks, thereby improving overall computation speed while maintaining grasp quality.
2Productivity
If existing grasp generation techniques are used, then grasps can be computed, but simplifications result in low grasp quality
Solution Approach 1:
The method performs preliminary grasp generation using simplified models before applying computationally intensive collision detection. By pre-filtering candidates using faster approximation methods, the system reduces the number of grasps that require expensive detailed collision checks, thereby improving overall computation speed while maintaining grasp quality.
Solution Approach 2:
The patent replaces complex mechanical collision detection with a signed distance field (SDF) based approach. The SDF provides a continuous, differentiable representation of collision risk that can be efficiently computed and optimized, substituting traditional discrete mechanical contact models with a smoother mathematical field that enables faster computation and better grasp quality assessment.
3Reliability
If collision prediction is used to filter grasp candidates, then collision-free grasps are selected, but grasps with minor collisions are unnecessarily ruled out
Solution Approach 1:
The optimization process applies beforehand cushioning by adjusting grasp parameters to account for expected part movements during grasping. The system anticipates that parts may shift slightly when grasped and optimizes grasps to maintain quality even with these movements, thereby accepting grasps that would be rejected by strict static collision checks but would work well in practice.
Solution Approach 2:
The patent creates a signed distance field as a continuous mathematical copy of the collision landscape. Instead of using discrete binary collision indicators, the SDF provides a smooth gradient field that represents collision risk continuously, allowing the optimizer to find grasps that minimize collision risk while accommodating realistic part movements and deformations.
Data Source
AI summary
A robotic grasp generation technique for part picking applications. Part and gripper geometry are provided as inputs, typically from CAD files. Gripper kinematics are also defined as an input. A set of candidate grasps is provided using any known preliminary grasp generation tool. A point model of the part and a model of the gripper contact surfaces with a clearance margin are used in an optimization computation applied to each of the candidate grasps, resulting in an adjusted grasp database. The adjusted grasps optimize grasp quality using a virtual gripper surface, which positions the actual gripper surface a small distance away from the part. A signed distance field calculation is then performed on each of the adjusted grasps, and those with any collision between the gripper and the part are discarded. The resulting grasp database includes high quality collision-free grasps for use in a robotic part pick-and-place operation.