No abstract

There has been a growing interest in developing data-driven, in particular deep neural network (DNN) based methods for modern communication tasks. For a few popular tasks such as power control, beamforming, and MIMO detection, these methods achieve state-ofthe-art performance while requiring less computational efforts, less channel state information (CSI), etc. However, it is often challenging for these approaches to learn in a dynamic environment where parameters such as CSIs keep changing.This work develops a methodology that enables data-driven methods to continuously learn and optimize in a dynamic environment. Specifically, we consider an "episodically dynamic" setting where the environment changes in "episodes", and in each episode the environment is stationary. We propose a continual learning (CL) framework for wireless systems, which can incrementally adapt the learning models to the new episodes, without forgetting models learned from the previous episodes. Our design is based on a novel min-max formulation which ensures certain "fairness" across different episodes. Finally, we demonstrate the effectiveness of the CL approach by customizing it to a popular DNN based model for power control, and testing using both synthetic and real data.

No abstract

Integrated sensing and communication (ISAC) has been envisioned as a key enabler in the next-generation wireless networks. In this paper, we consider the joint information and sensing beamforming design in a multi-user and multi-target multi-input multi-output ISAC system, where a transmit BS and a sensing BS collaborate to sense targets, and the transmit BS sends information streams to communication users at the same time. To optimize the sensing performance and guarantee the communication throughput, we formulate a joint beamforming design problem to minimize the trace of the weighted Cramer–Rao bound of target parameters subject to the sum-rate constraint. The problem is challenging to solve due to the intricate non-convex objective function and constraints. We firstly exploit the weighted mean square error minimization (WMMSE) and semidefinite relaxation (SDR) techniques to devise a WMMSE–SDR algorithm that can achieve a KKT point of the problem. The SDR can be shown to be tight for a subproblem in the WMMSE–SDR algorithm, which implies zero duality for the subproblem. Based on this property and fractional programming techniques, we further reformulate the beamforming problem as a min–max form with simple constraints which then can be efficiently solved by first-order min–max optimization algorithms. Finally, the proposed algorithms are evaluated extensively in simulations. Numerical results show that both proposed algorithms can achieve promising performance in sensing and communication, and the low-complexity algorithm has a significantly reduced computation time.