A Scalable and Adaptive Approach to Data Fusion for Autonomous UAVs

Unlike standard Level 1 data fusion, advanced situational awareness and predictive battlespace awareness technologies must operate on a global view of the battlespace. To handle the complexity of an advanced data fusion system, we propose a quadtree-based architecture. This approach breaks down the virtual battlespace into many smaller subspaces, of varying sizes. The subspaces are organized in a tree structure where the root node represents the entire virtual battlespace. Nodes in lower levels of the quadtree represent smaller regions of the virtual battlespace. Each node performs computations on the tracks that reside in the area of the battlespace that it is responsible for. Each node then passes its fused tracks (and other related information) to its parent node in the quadtree. The quadtree-based architecture allows data fusion to be processed in a hierarchal, and efficient, manner. Tracks are fused at the lowest level of the tree possible, reducing the workload of the higher-level nodes. The independent nature of the nodes in the architecture allows the solution to be scalable and easily distributed. As a side benefit, distributing the nodes closer to the “data consumers” reduces overall communication cost.