Robot Sensor Fusion Confidence Control for Faulty Measurements
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sensor fusion methods in robotic control are susceptible to deterioration in measurement quality from individual sensors, which can adversely affect the overall fusion result, particularly in multipath architectures where poor measurement quality can impact multiple modalities.
Innovation Solution
A method using a neural network to determine confidence information for position predictions by analyzing distances between sensor data combinations, allowing the system to identify and mitigate the impact of poor sensor measurement quality through permutation-invariant machine learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sensor data fusion is performed using multiple sensors and modalities, then the reliability of object detection is improved, but the system becomes vulnerable to deterioration in measurement quality from individual sensors affecting multiple modalities
Solution Approach 1:
The patent segments the sensor fusion process into multiple independent modalities, where each modality processes sensor data through separate neural network pathways. This segmentation allows the system to identify and isolate poor quality measurements from individual sensors or sensor combinations, preventing them from degrading the entire fusion result. Each modality can be independently evaluated and weighted based on its confidence level.
Solution Approach 2:
The patent implements a feedback mechanism where the system determines confidence levels for each modality based on the agreement between multiple sensor combinations. This confidence information feeds back into the fusion process, allowing the system to dynamically adjust the weighting of different modalities. Modalities with low confidence (indicating poor sensor quality) are automatically downweighted or excluded, while high-confidence modalities receive greater weight in the final result.
2Reliability
If redundant sensor combinations are used to detect sensor failures, then the system can identify poor measurement quality, but the device complexity increases
Solution Approach 1:
The patent employs universal neural network architectures that can process multiple sensor types and combinations through the same computational framework. The same neural network modules and confidence determination mechanisms are applied across all modalities, regardless of which specific sensors are combined. This multi-functionality allows the system to handle various sensor configurations without requiring separate specialized processing paths for each combination, thereby reducing overall complexity.
Solution Approach 2:
The patent changes the parameter space by representing sensor combinations and their quality assessments in a standardized mathematical framework. Confidence levels are expressed as numerical parameters that can be directly compared and weighted, rather than requiring complex qualitative analysis of each sensor combination. This parameter transformation simplifies the integration of multiple sensor modalities and the determination of their relative reliability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
According to various embodiments, a method is described for controlling a robot device, including receiving, from each sensor of a plurality of sensors, a respective sensor data set from the sensor; determining, for each object of a set of objects containing at least one object, for each of a plurality of different combinations of the sensor data sets, a position prediction for the object by means of sensor data fusion of the sensor data sets according to the combination of the sensor data sets; determining, for each object of the set of objects, for each pair of a plurality of pairs of combinations, a distance between the position predictions determined for the object according to the combinations of the pair; feeding the determined distances to a neural network trained to determine confidence information for the position predictions from distances between position predictions for the pairs of combinations; and controlling the robot device using one or more of the position predictions taking into account the confidence information.