With the rapid development of Internet of Things (IoT) technology, the number of IoT devices has increased dramatically and a large amount of data has been generated. In order to further reduce the resource cost required for task offloading, it is necessary to design task offloading methods with high-energy efficiency and low latency. Considering the correlation between task offloading process and time in real-time interactive scenarios, we propose an evolutionary algorithm (EA) framework with online load prediction based on CNN-GRU hybrid model and channel attention mechanism (AM). In the model construction stage, we first combined convolutional neural network (CNN) and gated recurrent unit (GRU) to learn the features and patterns of historical data. In order to reduce the loss of historical information, the channel AM is introduced into the CNN-GRU model to enhance the influence of important features between information. In the model training stage, the optimal individual training model generated by the EA is used to further optimize the training accuracy and training effect of CNN-GRU-AM. In the test phase, the optimized CNN-GRU-AM network is used to predict the task load online and dynamically allocate computing resources while training the model online iteratively, which further reduces the delay and energy consumption of the task and improves the offloading performance of the system. The simulation results show that the proposed algorithm effectively reduces the system delay and the overall energy consumption.

Mobile Edge Computing (MEC) plays a pivotal role in optimizing the Industrial Internet of Things (IIoT), where the Industrial Task Offloading Problem (ITOP) is crucial for ensuring optimal system performance by balancing conflicting objectives such as delay, energy consumption, and cost. However, existing approaches often oversimplify multi-objective optimization by aggregating conflicting goals into a single objective, while also suffering from limited exploration and robustness in uncertain MEC scenarios within IIoT. To overcome this limitation, we propose EMDRL-ITOP, an Evolutionary Multi-Objective Deep Reinforcement Learning algorithm that synergizes an evolutionary algorithm with deep reinforcement learning (DRL). Firstly, we formulate a multi-objective task scheduling model for IIoT-MEC and design a three-dimensional vector reward function within a Multi-Objective Markov Decision Process framework, enabling simultaneous optimization of delay, energy, and cost. Then, EMDRL-ITOP integrates evolutionary mechanisms to enhance exploration and robustness: a dynamic elite selection strategy prioritizes high-quality policies, a distillation crossover operator fuses advantageous traits from elite strategies, and a proximal mutation mechanism maintains population diversity. These components collectively improve learning efficiency and solution quality in dynamic environments. Extensive simulations across six instances demonstrate that EMDRL-ITOP achieves a superior balance among conflicting objectives compared to state-of-the-art methods, while also outperforming existing algorithms in several key performance metrics.

The development of Industrial Internet of Things (IIoT) has completely changed traditional manufacturing industry. Industrial equipments with limited resources often cannot meet the diverse demands of numerous computing-intensive and latency-sensitive tasks. Mobile edge computing (MEC) offloads these tasks to nearby edge servers to achieve lower latency and energy consumption. However, considering the channel interferences of the network and the diverse demands of different tasks, coordinating computation offloading among multiple devices is challenging. To address this challenge, the computation offloading is formulated as a multiobjective optimization problem, and a new task model composed of scientific workflow tasks and concurrency workflow tasks is proposed to represent the multitask in the industrial environment. In addition, a two-hierarchical optimization framework is devised to optimize the bandwidth allocation and the multitask computation offloading through the dynamic bandwidth preallocation and the improved multiobjective evolutionary algorithm based on decomposition with two performance enhancing schemes. Comprehensive experiments demonstrate that the effectiveness and efficiency of our proposed framework in terms of the tradeoffs between latency and energy consumption, as well as the convergence and diversity of obtained nondominated solutions.

With the advancement of Internet of Things (IoT) technology, IoT smart devices (ISDs) are often frequently constrained by their battery capability and computing ability. In mobile edge computing (MEC) scenarios, ISDs increases computing capability while reducing energy consumption by offloading computation-intensive and latency-critical processes to edge servers. But the process of computing offloading (CO) leads to extra transmission time, which is unacceptable for IoT applications. Most current studies regard the CO as a multiobjective optimization (MOO) problem involving application completion time and energy consumption. However, with the increasing number of ISDs and the complexity of ISDs-based applications, it becomes extremely challenging to acquire the best solution through MOO. Evolutionary multitasking optimization (EMTO), as a novel search paradigm in evolutionary computation, utilizes intertask correlation and facilitates the solution of multiple tasks simultaneously. Therefore, we model the CO of ISDs as a multitasking multiobjective CO problem (MMCOP) by constructing an auxiliary task similar to the target task for the first time. In addition, we put forward a multitasking MOO algorithm based on reinforcement learning (RLMTO-CO) to solve MMCOP. Each task has a knowledge transfer probability value adaptively adjusted by reinforcement learning (RL), and the generation of auxiliary tasks is updated adaptively. The performance of RLMTO-CO has been validated in different test cases close to true IoT environments, and the experimental results show the superior competitiveness.

Unmanned aerial vehicle (UAV) and base station jointly assisted multiaccess edge computing (UB-MEC) technology is a promising direction to provide flexible computing services for resource-limited devices. Due to the non-real-time observation of device loads and the dynamic nature of demand in UB-MEC, it is a highly challenging problem to make UAV respond in real time to meet user’s dynamic preferences in UB-MEC. To this end, we propose a multiobjective deep reinforcement learning (MODRL) for computation offloading and trajectory control (COTC) of UAV. First, the problem is formulated as a multiobjective Markov decision process (MOMDP), where the traditional scalar rewards are extended to vector, corresponding to the number of task data collected, the completion delay, and the UAV’s energy consumption, and the weights are dynamically adjusted to meet different user preferences. Then, considering the device load information stored in UAV is non-real-time, an attentional long short-term memory (ALSTM) network is designed to predict real-time states by autofocusing important historical information. The near on-policy experience replay (NOER) reviews experiences close to on-policy can better promote learning of current strategy. The simulation results show that the proposed algorithm can obtain the action policy which meets the user’s time-varying preferences, and can achieve a good balance between different objectives under different preferences.

With the development of Internet of Vehicles (IOV) technology, minimizing network latency while performing computationally intensive tasks has become a challenge. As the most advanced data communication technology, 6G can significantly reduce the latency. Additionally, utilizing remote satellites for assistance can alleviate the computational pressure. In this paper, an Integrated Satellite-Terrestrial Internet of Vehicles (IST-IOV) environment was established with 6G networks. Firstly, we established the Edge Computation Service Provider (ECSP) to manage task allocation and data flow. Secondly, considering energy consumption, latency, and offloading process, we define utility values for vehicles and ECSP, construct a Stackelberg bi-level dynamic cooperative game. The existence of the unique Nash equilibrium in the game has also been proved by backward induction methods. Finally, based on the interaction between vehicles and ECSP, we proposed a computation offloading algorithm called Dynamic Cooperative Game Theory in IOV (DCGT-IOV). The DCGT-IOV algorithm can find the optimal offloading strategy for vehicles and the optimal pricing strategy for ECSP based on the gradient of the utility function. Simulation results demonstrate that the proposed algorithm achieves higher overall utility in various scenarios compared with other strategies and exhibits good convergence properties.

Nowadays, Internet of Vehicles plays an important role in the emerging intelligent transportation systems. In Internet of Vehicles, it is crucial to make appropriate temporal data scheduling to ensure the service quality and the high service ratio to meet the requests from vehicles. However, service quality and service ratio are two conflict goals because of the limited bandwidth and the vehicle mobility in Internet of Vehicles. In order to optimize these two conflict objectives simultaneously, we present an improved decomposition based multi-objective evolutionary algorithm for the temporal data scheduling (I-MOEA/D-TDS) in Internet of Vehicles. Based on the MOEA/D framework, we integrate a self-adaptive weight vector adjustment method based on chain segmentation to improve the performances of temporal data scheduling. For verifying the availability of presented algorithm, under the hybrid Vehicle-to-Infrastructure / Vehicle-to-Vehicle (V2I/V2V) communications and multiple Roadside Units (RSUs) scenario, we compare the proposed algorithm with several related algorithms under the effects of data valid periods, service workloads, maximum tolerated delay, and traffic workloads. Experimental results suggest that the presented algorithm can achieve better data service quality and service ratio.

Internet of Things (IoT) aims to provide ubiquitous services in real life. When different service requests arrive, how to assign them to proper service providers has become a challenging problem, especially in large-scale IoT service circumstances. In order to obtain the best service matching scheme, it is crucial to minimize total service cost and service time. Since both goals are conflicting, we have modeled IoT service as a multiobjective problem. Thus, we propose an improved decomposition-based multiobjective evolutionary algorithm for the IoT service (I-MOEA/D-IoTS). We have designed appropriate operators, such as array encoding, population initialization, Tchebycheff decomposition approach, local improvement, simulated binary crossover, and Gaussian mutation. In order to verify the effectiveness of the proposed algorithm, we apply it in three different scenarios of the agricultural IoT service. The simulation experimental results show that the proposed algorithm can achieve better tradeoff of solutions for IoT service and reduce total service cost and shorten service time.