AMR Object Segmentation for Payload Contact and Obstruction Avoidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current obstruction detection systems in autonomous mobile robots (AMRs) cannot differentiate between objects that the AMR is permitted to interact with and those it should avoid, leading to potential collisions when approaching payloads.
Innovation Solution
An object segmentation system that uses point cloud data, payload pose, and semantic data to segment detected objects into obstructions and allowed objects, allowing the AMR to interact with specific payloads while avoiding other objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If obstruction detection systems are turned off when the AMR approaches a payload, then the AMR can make contact and engage with the payload, but the AMR risks making contact with other objects that should be avoided
Solution Approach 1:
The system segments the point cloud data into different categories: payload points, obstruction points, and allowed object points. This segmentation allows the AMR to differentiate between objects that should be avoided (obstructions) and objects that are safe to interact with (allowed objects), resolving the contradiction by enabling simultaneous collision avoidance and payload interaction capability
Solution Approach 2:
The system introduces an intermediary classification layer between the sensor data and the navigation system. This intermediary classifies detected objects into obstructions and allowed objects based on point cloud analysis, enabling the AMR to maintain obstruction detection active while safely interacting with payloads and other allowed objects
2Reliability
If obstruction detection systems remain on when the AMR approaches a payload, then the AMR avoids collisions with obstructions, but the AMR cannot make physical contact and engagement with the payload
Solution Approach 1:
The system segments detected objects into obstructions and allowed objects using point cloud analysis. Payloads are classified as allowed objects rather than obstructions, enabling the obstruction detection system to remain active while permitting safe interaction with payloads. This segmentation resolves the contradiction by allowing differential treatment of different object types
Solution Approach 2:
The system applies different quality attributes to different regions of the environment. Payload regions are marked with 'allowed' quality attributes while obstruction regions are marked with 'avoid' attributes. This local quality differentiation enables the AMR to navigate with collision avoidance active while selectively allowing contact with specific objects based on their classified quality
3Device complexity
If the AMR uses a general obstruction detection system, then the system is simple to implement, but it cannot differentiate between objects to interact with and objects to avoid
Solution Approach 1:
The system segments point cloud data into distinct object categories using semantic information and spatial analysis. By dividing the detected environment into obstructions, payloads, and other allowed objects, the system gains differentiation capability while building upon the existing simple obstruction detection framework, thus resolving the contradiction between simplicity and adaptability
Solution Approach 2:
The system changes the parameters used for object classification by incorporating semantic data and spatial relationships into the detection algorithm. This allows the existing detection system to differentiate between object types by modifying classification parameters rather than requiring a completely new complex system
Data Source
AI summary
In accordance with one aspect of the inventive concepts, provided is an autonomous mobile robot, comprising: at least one processor in communication with at least one computer memory device; at least one sensor configured to acquire point cloud data; a pallet detection system configured to provide a pose of a payload; and an object segmentation system comprising computer program code executable by the at least one processor to segment detected objects into obstructions and allowed objects based on the point cloud data, the pose of the payload and semantic data about the payload. A corresponding method is also provided.