Abstract:Although multi-access edge computing (MEC) has allowed for computation offloading at the network edge, weak wireless signals in the radio access network caused by obstacles and high network load are still preventing efficient edge computation offloading, especially for user requests with stringent latency and reliability requirements. Intelligent reflective surfaces (IRS) have recently emerged as a technology capable of enhancing the quality of the signals in the radio access network, where passive reflecting elements can be tuned to improve the uplink or downlink signals. Harnessing the IRS's potential in enhancing the performance of edge computation offloading, in this paper, we study the optimized use of a system of multi-IRS along with the design of the offloading (to an edge with multi MECs) and resource allocation parameters for the purpose of minimizing the devices' energy consumption considering 5G services with stringent latency and reliability requirements. After presenting our non-convex mathematical problem, we propose a suboptimal solution based on alternating optimization where we divide the problem into sub-problems which are then solved separately. Specifically, the offloading decision is solved through a matching game algorithm, and then the IRS phase shifts and resource allocation optimizations are solved in an alternating fashion using the Difference of Convex approach. The obtained results demonstrate the gains both in energy and network resources and highlight the IRS's influence on the design of the MEC parameters.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is how to optimize task offloading decisions and resource allocation in a multi - intelligent reflecting surface (Multi - IRS) - assisted mobile edge computing (MEC) system to meet the requirements of high - reliability and low - latency services while minimizing the energy consumption of devices. Specifically, the paper focuses on improving the performance of edge - computing offloading by using multiple IRSs to improve wireless communication quality in 5G and future networks. The main objectives of the paper are:
1. **Optimize task offloading decisions**: Determine which tasks of user equipment (UE) should be offloaded to which edge servers (MEC) and how to allocate computing resources to meet the latency and reliability requirements of tasks.
2. **Optimize IRS phase shift**: Adjust the phase shift of the reflective elements on the IRS to enhance signal quality and reduce transmission latency and energy consumption.
3. **Optimize resource allocation**: Reasonably allocate resource blocks (RB) of the orthogonal frequency - division multiple - access (OFDMA) system and the transmission power of UEs to further reduce energy consumption.
### Paper Background
- **Multi - access edge computing (MEC)**: Although MEC technology has already allowed computing offloading at the network edge, poor radio access network (RAN) communication quality due to weak wireless signals (such as obstruction blocking, high network load, etc.) still hinders efficient edge - computing offloading, especially for user requests with strict latency and reliability requirements.
- **Intelligent reflecting surface (IRS)**: A recently emerged technology that improves the reliability of wireless communication by adjusting passive reflective elements to improve the uplink or downlink signal quality.
### Main Contributions
1. **Joint optimization problem modeling**: Propose a non - convex mixed - integer nonlinear programming (MINLP) model to jointly optimize the phase shift of IRS elements, offloading decisions, the transmission power of UEs, OFDMA resource block allocation, and server computing resource allocation while ensuring the latency and reliability required for services and minimizing the total offloading energy consumption.
2. **Alternate optimization method**: Propose an alternate - optimization - based method to decompose the problem into multiple sub - problems for separate solution. First, use the matching game algorithm to solve the offloading decision problem, and then iteratively solve the IRS phase - shift and resource - allocation parameters to finally achieve target convergence.
3. **Numerical analysis**: Demonstrate the performance of solutions under different system configurations through numerical results and study in detail the impact of optimizing IRS elements on reducing UE energy consumption.
### Mathematical Model
- **Transmission rate**:
\[
R_{k,j,b}(p_k, \theta_j) = W\log_2\left(1 + \frac{p_k|h_{k,j,b}+h^I_{k,j,b}\Theta_jg_{j,b}|^2}{WN_0}\right)
\]
where \(p_k\) is the transmission power of UE \(k\), \(W\) is the resource - block bandwidth, and \(N_0\) is the white - noise power level.
- **Upload latency**:
\[
L^u_{k,j}(y_k, p_k, \theta_j)=\frac{D_k}{\sum_{b\in B}y_{k,b}R_{k,j,b}(p_k, \theta_j)}
\]
- **Computation latency**:
\[
L^c_{k,j}(f_{k,j})=\frac{D_kC_k}{f_{k,j}F_j}
\]
- **Reliability constraint**:
\[
1-\prod_{j\in M}(1 - x_{k,j}\phi_j)\geq\bar{R}_k
\]
### Solutions
1. **Offloading decision optimization**: Use the matching game algorithm to solve the offloading decision problem, ensuring that each UE's task is assigned to an appropriate access point (AP) and IRS while meeting the latency and reliability requirements.
2. **IRS phase - shift and resource - allocation optimization**: Through the alternate - optimization method, solve the IRS phase - shift and resource - allocation parameters respectively to finally achieve the goal of minimizing energy consumption.
Through these methods, the paper aims to provide an effective solution to achieve high performance in a multi - IRS - assisted MEC system.