Height-Map Grasp Planning for Unknown Object Handling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Robotic systems face challenges in grasping and manipulating objects due to their inability to effectively adjust grip based on touch, especially when dealing with diverse and unknown objects, which limits their flexibility and efficiency in tasks like sort-and-grasp operations.
Innovation Solution
A system and method for grasp execution that includes determining a height map, segmenting it, generating a set of proposed grasps, and executing a grasp, using a sensing system and a grasping manipulator, which allows for object-agnostic grasp planning and does not require objects to be known or similar to known objects, enabling flexible grasp planning and execution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional robotic systems use pre-programmed grasp patterns for known objects, then grasp execution is reliable for those specific objects, but the system cannot handle diverse or unknown objects
Solution Approach 1:
The height map is segmented into multiple layers representing different height ranges. The system generates grasp candidates for each layer separately, allowing systematic handling of objects at various heights while maintaining computational tractability through divided processing
Solution Approach 2:
The system transitions from traditional 2D image-based grasp planning to 3D height map-based planning. By incorporating height information and creating layered representations, the system adds a vertical dimension to grasp planning, enabling better handling of objects at different elevations and improving adaptability to diverse object configurations
2Adaptability or versatility
If robotic systems require extensive training data for specific object types, then grasp accuracy improves for those objects, but the system loses flexibility when encountering new object types
Solution Approach 1:
The system performs self-service by automatically generating grasp candidates from height map segmentation without requiring external training data or classification of object types. The algorithm inherently adapts to any object configuration by analyzing the geometric structure directly from sensor data, eliminating the need for pre-training on specific object categories
Solution Approach 2:
The system changes the fundamental parameter from object-type classification to height-based layering. Instead of adapting to different object categories through training, the system adapts to any object by segmenting the height map and generating appropriate grasp candidates for each layer, transforming the adaptation mechanism from learning-based to geometry-based
3Reliability
If the system generates comprehensive grasp candidates for all possible objects, then grasp coverage is complete, but computational resources are wasted on infeasible grasps
Solution Approach 1:
By segmenting the height map into discrete layers, the system limits grasp candidate generation to only those positions within each layer's height range. This segmentation approach ensures computational resources are focused on feasible grasps for objects at detectable heights, eliminating waste on infeasible candidates while maintaining comprehensive coverage of all possible object locations
Solution Approach 2:
The system generates grasp candidates for each height layer independently, performing partial action for each layer rather than attempting to generate all possible grasps simultaneously. This approach provides sufficient grasp coverage for objects at any height while avoiding the excessive computational burden of generating and evaluating all theoretically possible grasps in the entire workspace
Data Source
AI summary
Systems and method for grasp execution using height maps.