Joint task inference, which fully utilizes end edge cloud cooperation, can effectively enhance the performance of deep neural network (DNN) inference services in the industrial internet of things (IIoT) applications. In this paper, we propose a novel joint resource management scheme for a multi task and multi service scenario consisting of multiple sensors, a cloud server, and a base station equipped with an edge server . A time slotted system model is proposed, incorporating DNN deployment, data size control, task offloading, computing resource allocation, and wireless channel allocation. Among them, the DNN deployment is to deploy proper DNNs on the edge server under its total resource constraint, and the data size control is to make trade off between task inference accuracy and task transmission delay through changing task da ta size. Our goal is to minimize the total cost including total task processing delay and total error inference penalty while guaranteeing long term task queue stability and all task inference accuracy requirements. Leveraging the Lyapunov optimization, we first transform the optimization problem into a deterministic problem for each time slot. Then, a deep deterministic policy gradient (DDPG) based deep reinforcement learning (DRL) algorithm is designed to provide the near optimal solution. We further desi gn a fast numerical method for the data size control sub problem to reduce the training complexity of the DRL model, and design a penalty mechanism to prevent frequent optimizations of DNN deployment. Extensive experiments are conducted by varying differen t crucial parameters. The superiority of our scheme is demonstrated in comparison with 3 other schemes.
10 January 2023 in IEEE Transactions on Vehicular Technology
Edge computing is essential to enhance delay-sensitive and computation-intensive machine learning (ML) task inference services. Quality of inference results, which is mainly impacted by the task data and ML models, is an important indicator impacting the system performance. In this paper, we consider a quality-aware edge-assisted ML task inference scenario and propose a resource management scheme to minimize the total task processing delay while guaranteeing the stability of all the task queues and the inference accuracy requirements of all the tasks. In our scheme, the task offloading, task data adjustment, computing resource allocation, and wireless channel allocation are jointly optimized. The Lyapunov optimization technique is adopted to transform the original optimization problem into a deterministic problem for each time slot. Considering the high complexity of the optimization problem, we design an algorithm that decomposes the problem into a task offloading and channel allocation (TOCA) sub-problem, a task data adjustment sub-problem, and a computing resource allocation sub-problem, and then solves them iteratively. A low-complexity heuristic algorithm is also designed to solve the TOCA sub-problem efficiently. Extensive simulations are conducted by varying different crucial parameters. The results demonstrate the superiority of our scheme in comparison with 4 other schemes.
13 December 2023 in IEEE Internet of Things Journal
The combination of edge computing and blockchain can enhance task processing while ensuring security and credibility. To maximize system performance and avoid resource waste in task offloading, it is essential to match the resource allocation of the task computing and the blockchain consensus process. However, the existing works treated the above two processes as two independent processes and optimized them separately and ignored the above matching problem. In this paper, we propose a resource management scheme for blockchain-assisted edge computing (EC) networks consisting of multiple devices, multiple base stations equipped with edge servers, a cloud server, and a network controller deployed on the edge layer. To minimize the total task processing delay and energy consumption of the devices, we formulate a joint task processing problem incorporating task scheduling, transmit power control, and computing resource allocation. To match the computing delay and consensus delay of each task, we balance the computing resources allocated for the two processes. We design a deep reinforcement learning (DRL) algorithm that utilizes the twin delayed deep deterministic policy gradient (TD3) technology embedded with a fast numerical method, which effectively reduces the training complexity of the DRL model. Extensive experiments are conducted by varying 4 crucial parameters. The superiority of our scheme is demonstrated in comparison with 3 other reference schemes. The performance of our scheme is about 18.3%~24.1% higher than that of other schemes.
29 April 2026 in IEEE Transactions on Cloud Computing
In a blockchain-assisted edge intelligence system, blockchain can effectively enhance the reliability and trustworthiness of artificial intelligence (AI) task inference. When delegating an AI task from a user device to an edge server, blockchain will broadcast it to the entire network by recording it as a transaction and then conduct the verification process. Computing resource matching is a core issue of blockchain-assisted edge intelligence system, because block mining and task inference will face sharing of computing resources. The prerequisite for achieving optimal system performance is to match the block generation speed in block mining with the number of AI tasks per unit time in the edge intelligence system by computing resource allocation. However, prior studies have neglected the matching issue mentioned above and have not incorporated discussions on the properties of AI tasks, i.e., task inference accuracy. We propose a strategy for task offloading and resource allocation in the blockchain-assisted edge intelligence system. To minimize both the overall delay of task processing and the energy consumption of the user devices, we propose a combined optimization problem that includes task offloading decision, transmission power control, task data size adjustment, and allocation of computing resource on the premise of satisfying the task inference accuracy constraint. To this end, we propose a deep reinforcement learning (DRL) algorithm that incorporates the twin delayed deep deterministic policy gradient (TD3) with a swift numerical approach, significantly lessening the time complexity of the algorithm. To demonstrate the superiority of our scheme, we analyze the complexity and convergence of our algorithm, and then undertake numerous simulation experiments and compare the performance with the other 3 reference schemes in 6 scenarios.