Joint Computation and Communication Resource Optimization for Beyond Diagonal UAV-IRS Empowered MEC Networks

Intelligent Reconfigurable Surfaces (IRS) are crucial for overcoming
challenges in coverage, capacity, and energy efficiency beyond 5G (B5G). The
classical IRS architecture, employing a diagonal phase shift matrix, hampers
effective passive beamforming manipulation. To unlock its full potential,
Beyond Diagonal IRS (BD-IRS or IRS 2.0) emerges as a revolutionary member,
transcending limitations of the diagonal IRS. This paper introduces BD-IRS
deployed on unmanned aerial vehicles (BD-IRS-UAV) in Mobile Edge Computing
(MEC) networks. Here, users offload tasks to the MEC server due to limited
resources and finite battery life. The objective is to minimize worst-case
system latency by optimizing BD-IRS-UAV deployment, local and edge
computational resource allocation, task segmentation, power allocation, and
received beamforming vector. The resulting non-convex/non-linear NP-hard
optimization problem is intricate, prompting division into two subproblems: 1)
BD-IRS-UAV deployment, local and edge computational resources, and task
segmentation, and 2) power allocation, received beamforming, and phase shift
design. Standard optimization methods efficiently solve each subproblem. Monte
Carlo simulations provide numerical results, comparing the proposed
BD-IRS-UAV-enabled MEC optimization framework with various benchmarks.
Performance evaluations include comparisons with fully-connected and
group-connected architectures, single-connected diagonal IRS, and binary
offloading, edge computation, fixed computation, and local computation
frameworks. Results show a 7.25% lower latency and a 17.77% improvement in data
rate with BD-IRS compared to conventional diagonal IRS systems, demonstrating
the effectiveness of the proposed optimization framework.