3D Vision Pick-and-Place Robot for Moving Bulk Objects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing pick and place robot systems struggle to efficiently handle randomly sized, shaped, and textured objects from a bulk with high throughput and accuracy, especially in sorter systems where objects are manually inducted due to the complexity of picking and placing objects from a continuously moving conveyor.
Innovation Solution
A robot system utilizing a combination of 3D imaging and control algorithms, including machine learning, to accurately select and orient objects for gripping, with a flexible gripper configuration and feedback mechanisms for continuous improvement, achieving high success rates and throughput.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If a robot system is used to pick objects from a moving bulk, then automation is improved, but picking reliability deteriorates due to random object positions and shapes
Solution Approach 1:
The system performs preliminary actions by capturing images of the bulk objects before the picking operation, processing these images to identify object positions and characteristics, and pre-calculating the picking strategy. This allows the robot to adapt to random object arrangements without compromising reliability, as the system is already prepared with object location and gripper configuration information before the actual picking attempt.
Solution Approach 2:
The system employs dynamic adaptation by adjusting the gripper configuration based on real-time image analysis of the bulk objects. The robot can modify its gripping strategy on-the-fly to accommodate random object positions, shapes, and orientations, maintaining high picking reliability while fully automated. The control system dynamically selects which object to pick next and how to grip it, rather than following a fixed predetermined sequence.
2Productivity
If the robot picks objects at high speed, then productivity is improved, but picking precision deteriorates
Solution Approach 1:
The system performs preliminary image capture and processing before the actual picking action, allowing high-speed operation without sacrificing precision. By pre-identifying object positions and characteristics from images taken upstream, the robot can execute precise picking movements at high speed without needing to slow down for real-time visual feedback during the critical picking moment.
Solution Approach 2:
The system replaces traditional mechanical vision systems with advanced image processing and control algorithms that can rapidly analyze object positions and calculate precise gripping parameters. This computational approach enables high-speed precision picking by performing complex visual-motor coordination through software rather than mechanical means, significantly increasing throughput while maintaining accuracy.
3Adaptability or versatility
If the gripper is designed for flexible configuration, then adaptability is improved, but device complexity increases
Solution Approach 1:
The system employs a dynamically adjustable gripper that can change its configuration based on the specific object being picked. The control system processes images to determine object characteristics and automatically adjusts gripper parameters such as finger position, opening width, and gripping force. This dynamic adaptability allows a single gripper design to handle diverse object types without requiring multiple specialized grippers, balancing versatility with manageable complexity through software control.
Data Source
Figure 1~2
Figure 3
Figure 4a~4b
AI summary
A robot system for picking randomly shaped and sized object from a continuously moving stream of objects in bulk, e.g. a 3D bulk, and placing the object singulated and aligned on an induction or directly on a sorter. A pick and place robot has a robotic actuator for moving a gripper with a controllable gripping configuration of its gripping members, e.g. four suction cups, to adapt the gripper for various objects. A control system processes a 3D image of objects upstream of a position of the pick and place robot, identifies separate objects in the 3D image, and selects which object to grip, based on parameters of the identified separate objects determined from the 3D image. Based on e.g. size and shape of the selected object to grip, the gripping configuration of the gripper is adjusted to match the surface of the object to grip for optimal gripping. The robotic actuator, e.g. a gantry type robotic actuator, is then controlled to move the gripper to a position for gripping the object, and afterwards move the gripper with the gripped object to a target position and with a target orientation to release grip of the object and thus place the object on an induction or directly on a sorter. An image after placing the object along with properties of the object determined from the 3D image can be used as input to a machine learning for online improving pick and place performance of the robot system, e.g. for online improving the algorithm for selection of which object to pick, and also for selection of the appropriate gripping configuration to match the object.