Digitally Modulated Radar (DIMORA)

The focus of this research project is on the optimal design of digital MIMO radar. In this context, we naturally think about the performance bounds of different waveforms in terms of resolution in range, Doppler and angular domains. Furthermore, we also aim at optimal radar design for antenna arrays as well as both the transmitter and the receiver. Finally, adaptivity of the design will be considered.

1. Theoretical bounds

Our optimal MIMO radar design will be guided by certain radar performance criteria. While many research results only use identifiability as a criterium, we want to focus on more specific performance criteria. For instance, target detection probability is a criterium we will investigate, where we can consider either a Neyman Pearson or a Bayesian setting. The actual detection probability might be hard to optimize, but several distances between the probability densities of the two hypotheses (target present or not) can be considered as a manageable proxy. Furthermore, parameter estimation criteria will be considered such as the Fisher information or the Cramér-Rao lower bound, e.g., for estimating the range, Doppler, angle and/or radar cross section. Such criteria will be investigated for static as well as dynamic scenarios, where the latter have the advantage that we can use relatively accurate prior knowledge from past estimates.

2. Optimal radar design

Once we have derived the relevant performance measures, we can start optimizing the radar sequence/frame/antenna structures in order to obtain the best performance under some constraints that limit the amount of resources we are allowed to use (hardware, power, space, etc.). Thereby we will take additional interferences into account, such as self-interference as well as interference from other radars. For a static scenario, we have only a very limited knowledge of the number of targets and their location (e.g., through some rough occupancy maps). For this scenario, we could then for instance fall back to a 2-target approach where we assume 1 target of interest and 1 interfering target that models the total interference term. For a dynamic scenario, however, we will consider a more accurate model since we can obtain a more informative prior from past estimates through a tracker.

3. Closed loop radar

For a dynamic environment, the earlier mentioned optimization should be carried out in an adaptive fashion leading to adaptive resource allocation on both the transmit and receive side. This will naturally require closing the loop, with in this loop both a tracker and a limited capacity feedback channel. We plan to not only select how to adapt the MIMO radar structure based on this feedback, but we will also optimize the type of information that should be sent over the feedback channel, thereby taking the limited capacity into account. Both these tasks will rely on dynamic online optimization theory.

Finally, we want to analyze the performance of all the developed methods. We will therefore focus on simulated yet realistic data.