Towards Smarter Sensing: 2D Clutter Mitigation in RL-Driven Cognitive MIMO Radar

TL;DR

The paper tackles multitarget detection under unknown 2D clutter in a 6G context by proposing a cognitive MIMO radar that uses SARSA-based reinforcement learning to adapt transmitted waveforms and beamforming alongside a robust Wald-type detector. The system employs planar uniform planar arrays (UPAs) and a 2D disturbance model built on AR/ARMA processes with heavy-tailed noise, enabling effective 2D beam steering and clutter suppression. Key contributions include the integration of SARSA for adaptive waveform/beamforming, a per-bin robust detection statistic with a $P_{FA}$-controlled threshold, and a demonstration that learning-based adaptation yields significant detection gains over omnidirectional transmission, particularly at low SNR and as the array size grows to $N= N_T N_R = 10^4$. This work extends 1D disturbance studies to 2D planar arrays, highlighting the practical impact for robust sensing in dynamic 6G environments.

Abstract

Motivated by the growing interest in integrated sensing and communication for 6th generation (6G) networks, this paper presents a cognitive Multiple-Input Multiple-Output (MIMO) radar system enhanced by reinforcement learning (RL) for robust multitarget detection in dynamic environments. The system employs a planar array configuration and adapts its transmitted waveforms and beamforming patterns to optimize detection performance in the presence of unknown two-dimensional (2D) disturbances. A robust Wald-type detector is integrated with a SARSA-based RL algorithm, enabling the radar to learn and adapt to complex clutter environments modeled by a 2D autoregressive process. Simulation results demonstrate significant improvements in detection probability compared to omnidirectional methods, particularly for low Signal-to-Noise Ratio (SNR) targets masked by clutter.

Figures (5)

• Figure 1: RL-based MIMO radar system with UPA forming beams for target tracking. The radar agent sends actions to the UPA, receiving feedback in the form of state and reward signals for adaptive tracking via reinforcement learning.
• Figure 2: Disturbance PSD along with targets locations as red circles.
• Figure 3: Detection performance of RL beamforming versus omnidirectional with equal power allocation under $P_{FA} = 10^{-5}$ and $N = N_TN_R = 10^4$.
• Figure 4: $P_D$ using RL and alternative approaches of existing targets across different virtual antenna array size
• Figure 5: $P_D$ using RL and alternative approaches over SNR for target at position $\nu_1 = (-0.4,-0.4)$