Collision Detection Model for False Positive Suppression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing collision avoidance systems for machines that change shape, such as digging or loading machines, struggle to distinguish between movable parts of the machine and foreign objects, leading to false positive collision notifications.
Innovation Solution
A collision detection system that includes proximity sensors, positioning sensors, and a controller with a model of the machine. The controller updates the machine model based on the pose of movable parts and compares detected objects with expected detection results to determine whether the detection is from a movable part or a foreign object.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the sensor field-of-view includes parts of the machine to account for them, then detection accuracy improves, but false positive collision notifications increase due to movable parts entering or leaving the FOV
Solution Approach 1:
The system dynamically updates the machine model using real-time pose information from positioning sensors to track movable parts. This allows the system to adapt to the changing shape and configuration of the machine, distinguishing between movable parts and foreign objects by comparing sensor detections against the updated model, thereby reducing false positives while maintaining detection accuracy
Solution Approach 2:
A machine model serves as an intermediary between the sensor detections and the collision avoidance logic. The model incorporates dimension information, sensor locations, and pose information to predict expected detections from movable parts. This intermediary layer filters out false positives by comparing actual sensor readings against model predictions, allowing the system to maintain an inclusive sensor FOV without suffering from false alarms
2Measurement precision
If a learned obstruction detection process is trained for each different machine, then detection accuracy improves, but system complexity and training cost increase
Solution Approach 1:
Instead of training a learned process for each machine, the system changes parameters by updating the machine model with current pose information from positioning sensors. This parameter-based approach allows the system to adapt to different machines and configurations without requiring complex training processes, maintaining detection accuracy while significantly reducing system complexity
Solution Approach 2:
The system creates a digital copy (machine model) of the physical machine that includes dimension information, sensor locations, and pose information. This model can be updated and used across different machine configurations without requiring physical training data collection, eliminating the need for costly and complex training processes while maintaining accurate detection
Data Source
AI summary
A collision detection system including a proximity sensor, positioning sensor, and controller. The controller includes a memory storing instructions, and a model of the machine that includes dimension information of the machine, location information of the proximity sensor, and pose information of movable portions of the machine. A processor is connected to the proximity sensor, positioning sensor, and memory, and is configured to update pose information of the model based on the pose of the movable portions determined using the positioning sensor. The processor may receive a detection of the object from the proximity sensor. The processor may determine an expected detection result of the proximity sensor using the model, and based on the location information of the proximity sensor and the updated pose information. The processor may compare the detection with the expected detection result to determine whether the detection results from the movable portion or a foreign object.