Distributed Joint Manifold Learning for Sensor Fusion Under Limited Bandwidth
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sensor fusion technologies face challenges in efficiently fusing high-dimensional, heterogeneous data from multiple sensor modalities with limited communication resources, leading to information loss and restrictive payload constraints in unmanned systems.
Innovation Solution
A distributed joint manifold learning (DJML) framework that enables nodes to form a joint manifold, determine optimum manifold learning algorithms, compute contributions, and update contribution tables based on local and neighbor nodes, facilitating decentralized data fusion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If centralized fusion framework using joint manifold learning algorithm is used to fuse high-dimensional heterogeneous sensor data, then data fusion accuracy is improved, but computing burden and system complexity increase significantly
Solution Approach 1:
The centralized fusion framework is segmented into multiple distributed nodes, each performing local manifold learning on its own sensor data. Instead of one central processor handling all high-dimensional data, the system divides the computation across numerous smaller nodes, reducing the computing burden on any single component while maintaining the ability to fuse heterogeneous data types through their respective manifold representations.
Solution Approach 2:
The patent transforms the problem from fusing high-dimensional raw sensor data directly to fusing lower-dimensional manifold representations. By mapping high-dimensional heterogeneous sensor data into lower-dimensional manifolds that preserve essential structure and relationships, the system reduces computational complexity while maintaining fusion accuracy.
2Adaptability or versatility
If more sensors are added to a single platform to increase data diversity, then sensor fusion capability is improved, but payload weight constraints are violated
Solution Approach 1:
Instead of concentrating multiple sensors on a single platform, the system segments the sensing function across multiple smaller platforms or nodes. Each node carries fewer sensors, respecting individual payload constraints, while the distributed network collectively achieves diverse multi-sensor fusion through the combination of different nodes' measurements.
3Productivity
If distributed paradigm is implemented to reduce computing burden, then system scalability is improved, but communication overhead increases
Solution Approach 1:
The patent extracts only the essential manifold representation from each node's local sensor data, rather than transmitting all raw high-dimensional sensor data to neighboring nodes. This extraction process removes redundant information while preserving the critical structural relationships, significantly reducing communication overhead while maintaining the benefits of distributed processing.
Data Source
AI summary
The present disclosure provide a system, a method, and a storage medium for distributed joint manifold learning (DJML) based heterogeneous sensor data fusion. The system includes a plurality of nodes; and each node includes at least one camera; one or more sensors; at least one memory configured to store program instructions; and at least one processor, when executing the program instructions, configured to obtain heterogeneous sensor data from the one or more sensors to form a joint manifold; determine one or more optimum manifold learning algorithms by evaluating a plurality of manifold learning algorithms based on the joint manifold; compute a contribution of the node based on the one or more optimum manifold learning algorithms; update a contribution table based on the contribution of the node and contributions received from one or more neighboring nodes; and broadcast the updated contribution table to the one or more neighboring nodes.