Quantum-assisted federated learning for radar-based object tracking in IoT-enabled environments

Object tracking with radar has become a key part of many IoT applications, such as smart transportation, autonomous robotics, and ambient surveillance. Nevertheless, conventional machine learning techniques have fatal problems, such as privacy, noisy radar signals, latency, and generalization in the distributed IoT systems. This paper proposes a new model, Quantum-Assisted Federated Learning (QAFL), a combination of quantum machine learning (QML) and federated learning (FL) to solve these problems, which can effectively and privately identify radar objects and their routes. The suggested QAFL architecture has a new Hybrid Angle-Amplitude Encoding (HAAE) scheme with multi-layer Variational Quantum Circuits (VQCs) to support the effective extraction of features in the presence of noisy and non-homogeneous radar sensor data. We also present a Quantum-Enhanced Federated Averaging (Quantum-FedAvg) algorithm that can be used to improve the efficiency of the training, privacy, and scalability of distributed IoT nodes. Comprehensive experimental tests based on the CARRADA automotive radar dataset show that QAFL achieves large performance improvements compared to classical federated learning baselines, in terms of classification accuracy improvements of up to 7.2 percentage points with the low signal-to-noise ratio (SNR) regime, trajectory prediction error reduction of up to 23 percent, as well as considerable communication overhead and training latency reductions. Such results highlight the enormous potential of quantum-enhanced federated learning systems to radar-based internet-of-things tracking systems.