Accelerated sparse signal decomposition for single-step estimation in multistatic radars and sensor networks
(2023)
Files
These-GillesMonnoyer.pdf
Open Access - Adobe PDF
- 7.29 MB
Details
- Authors
- Supervisors
- Abstract
- With advancements in electronics and computer science, new research possibilities have emerged, enabling the resolution of previously intractable complex problems. Notably, sensor network applications have benefited from this aspect regarding real-time feasibility for low-cost implementation. Built upon this observation, this thesis explores accelerations of robust sparsity-driven estimation techniques for sensor networks, with a particular focus on multistatic radars. The results of this research are divided into three acceleration contexts motivated by the requirements of radar applications and progressing from an abstract point of view toward application-specific considerations. First, we accelerate sparse signal decomposition algorithms aiming to solve challenging NP-hard sparse decomposition problems efficiently. Our novel approach accelerates standard Branch-and-Bound procedures solving the targeted decomposition problem. Second, we focus on decomposing continuous sparse signals in the estimation and sensing context. In particular, we propose a novel interpolation-based continuous version of the “Orthogonal Matching Pursuit” algorithm that is faster with respect to state-of-the-art propositions. We also provide extensions that are specifically tailored for radar systems. Lastly, we tackle “single-step” estimation processes in sensor networks by introducing a generalized acceleration framework for single-step methods called “Grid Hopping”. Bridging a gap between distinct research communities, we formally study existing trade-offs between (i) two-step procedures, commonly used for their simplicity and low computational requirements, and (ii) single-step methods offering more robust results but suffering from high computational complexity. The technique is ultimately particularized to the multistatic radar application and evaluated with numerical results and measurements obtained from our active multistatic radar system, tailored for this research. These assess the effectiveness of the acceleration provided by our propositions compared to existing algorithms.
- Affiliations
Citations
Monnoyer de Galland de Carnières, G. (2023). Accelerated sparse signal decomposition for single-step estimation in multistatic radars and sensor networks. https://hdl.handle.net/2078.5/236102