Scalable Probabilistic Data Association with Extended Objects
Florian Meyer, Zhenyu Liu, Moe Z. Win · 2019
Situation-aware technologies will create new services and applications in emerging fields such as autonomous transportation, home automation, and smart cities. A main challenge in situation-aware applications is data association with "extended" objects that originate an unknown number of measurements. To address this challenge, we introduce a new method based on factor graphs and the sum-product algorithm. Our approach makes it possible to reduce computational complexity in a principled manner due to an overcomplete description of data association uncertainty that enables a "stretching" of the factor graph. The complexity of the resulting sum-product algorithm for data association with extended objects (SPADA-X), only scales quadratically in the number of objects and linearly in the number of measurements. Without relying on heuristic preprocessing steps, it is suitable for localization and tracking of multiple objects that potentially generate a large number of measurements. Simulation results demonstrate the excellent estimation accuracy and scalability of SPADA-X.