Robotic Gripper Analytics for Overlapping Product Picking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Robotic gripper systems face challenges in accurately identifying and handling products that are touching or overlapping on a conveyor belt, leading to inefficiencies and potential damage due to unpredictable shifts and variability in product size, shape, and weight, which conventional systems struggle to adapt to.
Innovation Solution
Implementing a system that uses upstream sensors to image chaotic piles of products before they reach downstream robotic stations, coordinating with downstream sensors to re-image the pile as robotic arms make picks, and employing machine learning constructs for object discrimination, tracking, and grasp quality detection, with real-time adjustments to improve accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional sensors and control systems are used to identify products on a conveyor belt, then the system structure remains simple, but the system cannot accurately handle products that are touching or overlapping, leading to low pick efficacy
Solution Approach 1:
The vision system segments the chaotic pile of products into individual detectable units using machine learning algorithms. The system divides the complex task of identifying products in overlapping arrangements into smaller sub-tasks: detecting product boundaries, classifying product types, and tracking product positions across multiple frames. This segmentation enables accurate identification despite product contact and overlap.
Solution Approach 2:
The system performs preliminary imaging and analysis upstream before products reach the robotic picking station. By capturing images of the chaotic pile in advance and pre-processing the visual data to identify product locations, types, and relationships, the system prepares the information needed for accurate picking before the actual picking action occurs.
2Productivity
If the robotic system operates at high speed to achieve high throughput, then productivity increases, but the system cannot adapt to unpredictable product shifts and variability, leading to missed grasps and product damage
Solution Approach 1:
The system uses periodic imaging at high frame rates to continuously track product positions and detect shifts in the chaotic pile. By capturing images at regular intervals (e.g., 60+ frames per second), the system maintains real-time awareness of product locations and can adapt to unpredictable movements while operating at high picking speeds.
Solution Approach 2:
The system implements closed-loop feedback by continuously comparing actual product positions detected by the vision system with the planned picking trajectory. When product shifts or variability is detected, the controller adjusts the robotic gripper's position, orientation, and gripping force in real-time to compensate for deviations, ensuring reliable picking despite high-speed operation.
3Reliability
If the gripper applies high gripping force to prevent product drops, then product retention improves, but product damage increases due to variability in product size, shape, and weight
Solution Approach 1:
The system applies local quality by adjusting gripping parameters specifically for each detected product based on its identified characteristics. The vision system determines product size, shape, and material properties, and the controller configures the gripper's force distribution, contact points, and closure speed accordingly. This localized adaptation ensures adequate retention without excessive force that could cause damage.
Solution Approach 2:
The gripping force and gripper configuration are dynamically adjusted in real-time based on detected product properties. Rather than using a fixed gripping force, the system modifies the gripping parameters adaptively for each product instance, increasing force for heavier or smoother products and reducing force for lighter or more fragile products, thereby balancing retention and damage prevention.
4Measurement precision
If upstream imaging is performed to identify chaotic product arrangements, then pick accuracy improves, but processing time increases due to complex image analysis
Solution Approach 1:
The system performs preliminary imaging and basic processing upstream before products reach the picking zone. By capturing images early and pre-identifying product locations, types, and relationships in the chaotic pile, the system reduces the computational burden during the actual picking operation, enabling fast execution despite complex analysis requirements.
Solution Approach 2:
The system replaces traditional mechanical product presentation methods (which would require products to be neatly arranged) with a vision-based identification system. Instead of mechanically organizing products before picking, the system uses machine learning algorithms to automatically interpret chaotic arrangements from images, substituting computational processing for mechanical preparation and reducing overall processing time.
Data Source
AI summary
Exemplary embodiments pertain to techniques for generating and displaying a user interface for a robotic pick-and-place system. The user interface may include, among other features, a throughput of the pick-and-place system as reflected in analytics data determined as the system attempts to effect picks. A user can adjust parameters like the opening amount of the gripper or weights applied by filter and sort rules on the interface, and monitor how changes affect throughput.