A novel new waveform is compared with two other existing waveforms to create a radar network. In this radar network one primary radar transmits a waveform for sensing and communication and one or multiple radars receive these waveforms. The constraint for the radars is that all radars can build up a radar picture and situation awareness. What also means that the performance of the primary radar does not degrade. This new radar type is called cooperative hitchhiker with communications. In this radar network the main task is sensing, therefore additional communication signals are used to increase the performance of the sensing task. The advantages of parallel sensing and communications are reducing interference, dual use of the scarce electromagnetic (EM) spectrum in a congested EM environment and it is possible in a bistatic configuration to detect a bistatic scattering object.
The three waveforms which are compared are phase coded frequency modulated continuous wave (PC FMCW) linear frequency modulated – minimum shift keying (LFM-MSK) and a new waveform time delay between radar bursts (TDBRB). TDBRB is a simple but effective communication method on top of the sensing waveform and can be used with staggered waveforms. Where the first two waveforms make use of binary phase shift keying (BPSK), TDBRB uses a time modulation with the inserted time delay between two successive radar bursts. To distinguish the two radar bursts the first burst has a linear frequency modulated (LFM) upchirp and the successive burst, after the inserted time delay, has an LFM downchirp. PC FMCW and LFM-MSK data are compared with simulations and an experiment of the TDBRB waveform.
The conclusions are that PC FMCW has the highest data rate, followed by LFM-MSK, and TDBRB has the lowest data rate for line-of-sight connections. The first two make use of the information in one pulse, or FMCW chirp. This results in a lower signal-to-noise ratio (SNR) and is therefore only suited for line-of-sight connections. With a matched filter and coherent integration of the TDBRB signals, this waveform is most suited for non-line-of-sight connections and communication over an object-of-interest at the cost of a lower bit rate. TDBRB has a similar performance for the different Swerling cases as the detection probability of a regular radar. The degradation in performance with the TDBRB waveform is only a fraction of the burst time because the time delay is added to an original radar burst. These results make TDBRB most suited for non-line-of-sight communication and for communications with a low signal-to-noise ratio.

Persistent surveillance is an urgent proficiency. For security, surveillance cameras are a strong asset as they support the automatic tracking of people and are directly interpretable by a human operator. Radar on the other hand can be used under a broad range of circumstances: radar can penetrate mediums such as clouds, fogs, mist and snow, and it can be used when it gets dark.
However radar data, compared to an optical sensor as video, is not as easily interpretable by a human operator. This thesis explores the potential of multimodal deep learning with a radar and video sensor to improve the classification accuracy of human activity. A recorded and labelled dataset is created that contains three different human activities: walking, walking with a metal pole and walking with a backpack (10 kg). A Single Shot Detector is used to process the video data. The cropped frames are then associated with the start of a radar micro-Doppler signature with a duration of 1.28 seconds. The dataset is split in a training (80 %) and validation (20 %) set such that no data from a person in the training set is in the validation set. Implementations of convolutional neural networks for the video frames and micro-Doppler signatures obtain classification accuracies of 85.78 % and 63.12 % respectively for previously mentioned activities. It was not possible to distinguish a person walking and walking carrying a backpack on basis of the micro-Doppler signatures. The synchronised dataset is used to investigate different fusion methods. Both early and late fusion methods show an improvement in classification accuracy. The best obtained early fusion model achieves a classification accuracy of 90.60 %. Omitting the radar data however shows a drop in classification accuracy of just 0.9 %, identifying the video data as the dominant modality in this particular setup.