Radar systems have a plethora of civilian applications in automotive, surveillance, medicine among others due to their “beyond the sight” coverage and all-weather applicability. Many of these emergent applications like medical and automotive require high resolution to distinguish objects/ phenomena, while desiring a multi-view perspective. Further, targets are no longer desired to appear as “a blip on the radar”, but the system is required to provide an image like reconstruction of the scene. Furthermore, there is an ever-increasing requirement for high accuracy in parameter estimation as in industry automation. These call for a revisit of radar architectures, waveforms and receiver processing towards meeting the requirements. However, an analysis of the state-of-the-art indicates key shortcomings of fundamental importance towards design, optimization and implementation of such systems in diverse dynamic environments. Addressing these shortcomings necessitates addressing a number of challenges, not possible through incremental changes to the state-of-the-art. In fact, it calls for a systematic study building on key advances including significant contributions in terms of new radar architectures, waveforms and receiver processing to exploit spatial diversity in a wide-variety of problems. This motivates the pursued fundamental research on radar system design, optimization and adaptation in SPRINGER. Leveraging on developments in the communications, SPRINGER proposes a wide spectrum of radar architectures, from fully-coherent MIMO (similar to C-RAN) to decentralized architectures enabled by device to device communications in the emerging standards. Such a methodology provides for a system designer with a high resolution performance-complexity trade-off. In fully coherent MIMO, a centralized processing of all nodes ensures exploitation of diversity. In partially-coherent MIMO, a low rate feedback from central unit is assumed to enable regular updates; this is also a first step towards decentralization with significant processing at the nodes and fusion at the central node. In the decentralized MIMO, which mimicks a mesh network, there is no central unit and the nodes are assumed to communicate using evolving standards. For each of the architectures, waveform design, optimization and adaptation is pursued to exploit spatial diversity in extended targets based on the available information.An in-lab validation of the some of the algorithms considered will be taken up and a roadmap for a suitable candidate will be developed.