Edge Artificial Intelligence for Real-Time Target Monitoring

The key enabling technology for the exponentially growing cellular communications sector is location-based services. The need for location-aware services has increased along with the number of wireless and mobile devices. Estimation problems, and particularly parameter estimation, have drawn a lot of interest because of its relevance and engineers' ongoing need for higher performance. As applications expanded, a lot of interest was generated in the accurate assessment of temporal and spatial properties. In the thesis, two different approaches to subject monitoring are thoroughly addressed. For military applications, medical tracking, industrial workers, and providing location-based services to the mobile user community, which is always growing, this kind of activity is crucial. In-depth consideration is given to the viability of applying the Angle of Arrival (AoA) and Receiver Signal Strength Indication (RSSI) localization algorithms in real-world situations. We presented two prospective systems, discussed them, and presented specific assessments and tests. These systems were put to the test in diverse contexts (e.g., indoor, outdoor, in water...). The findings showed the localization capability, but because of the low-cost antenna we employed, this method is only practical up to a distance of roughly 150 meters. Consequently, depending on the use-case, this method may or may not be advantageous. An estimation algorithm that enhances the performance of the AoA technique was implemented on an edge device. Another approach was also considered. Radar sensors have shown to be durable in inclement weather and bad lighting conditions. Frequency Modulated Continuous Wave (FMCW) radars are the most frequently employed among the several sorts of radar technologies for these kinds of applications. Actually, this is because they are low-cost and can simultaneously provide range and Doppler data. In comparison to pulse and Ultra Wide Band (UWB) radar sensors, they also need a lower sample rate and a lower peak to average ratio. The system employs a cutting-edge surveillance method based on widely available FMCW radar technology. The data processing approach is built on an ad hoc-chain of different blocks that transforms data, extract features, and make a classification decision before cancelling clutters and leakage using a frame subtraction technique, applying DL algorithms to Range-Doppler (RD) maps, and adding a peak to cluster assignment step before tracking targets. In conclusion, the FMCW radar and DL technique for the RD maps performed well together for indoor use-cases. The aforementioned tests used an edge device and Infineon Technologies' Position2Go FMCW radar tool-set.