Deep Neural Network (DNN) has become an essential technology for edge intelligence. Due to significant resource and energy requirements for large-scale DNNs’ inference, executing them directly on energy-constrained Internet of Things (IoT) devices is impractical. DNN partitioning provides a feasible solution for this problem by offloading some DNN layers to execute on the edge server. However, the resources of edge servers are also typically limited. An energy-constrained and resource-constrained optimization problem is generated in such a realistic environment. Motivated by this, we investigate an optimization problem of DNN partitioning and offloading in a multiuser resource-constrained environment, which is considered an intractable Mixed-Integer Nonlinear Problem (MINLP). We decompose the problem into two subproblems and propose an Energy-Efficient DNN Partitioning and Offloading (EEDPO) strategy to solve it in polynomial time based on the minimum cut/maximum flow theorem and dynamic programming. Finally, we test the impact of energy constraint, DNN type, and device number on the performance of EEDPO. Simulation results on realistic DNN models demonstrate that the proposed strategy can significantly improve the DNN inference task completion rate compared to other methods.

No abstract

No abstract

With the development of smart Internet of Things (IoT), it has seen a surge in wireless devices deploying Deep Neural Network (DNN) models for real-time computing tasks. However, the inherent resource and energy constraints of wireless devices make local completion of real-time inference tasks impractical. DNN model partitioning can partition the DNN model and use edge servers to assist in completing DNN model inference tasks, but offloading also requires a lot of transmission energy consumption. Additionally, the complex structure of DNN models means partitioning and offloading across different network layers impacts overall energy consumption significantly, complicating the development of an optimal partitioning strategy. Furthermore, in certain application contexts, regular battery charging or replacement for smart IoT devices is impractical and environmentally harmful. The development of wireless energy transfer technology enables devices to obtain RF energy through wireless transmission to achieve sustainable power supply. Motivated by this, We propose a problem of joint DNN model partition and resource allocation in Wireless Powered Edge Computing (WPMEC). However, time-varying channel state in the WPMEC have a significant impact on resource allocation decisions. How to jointly optimize DNN model partition and resource allocation decisions is also a significant challenge. We propose an online algorithm based on Deep Reinforcement Learning (DRL) to solve the time allocation decision, simplifying a Mixed Integer Nonlinear Problem (MINLP) into a convex optimization problem. Our approach seeks to maximize the completion rate of DNN inference tasks within the constraints of time-varying wireless channel states and delay constraints. Simulation results show the exceptional performance of this algorithm in enhancing task completion rates.