This paper investigates the problem of multidimensional resource management in multi-access mobile edge computing (MEC) networks against external dynamic jamming. The objective is to minimize the long-term computational cost of the MEC network while satisfying the task computation delay requirements of user equipment (UE) by jointly optimizing computing and communication resource allocation. To overcome challenges such as frequency conflict and dynamic jamming attacks, a distributed multi-agent hierarchical deep reinforcement learning (MAHDRL) MEC framework based on hybrid heterogeneous decision-making is proposed. Specifically, a hierarchical MEC anti-jamming data offloading optimization model is constructed, and the MEC resource management problem is formulated as a decentralized partially observable Markov decision process (Dec-POMDP). Based on this, a distributed MAHDRL algorithm based on the actor-critic (AC) model is designed to solve the multi-agent high-dimensional nonlinear hybrid integer programming NP-hard problem: the high-level network in the base station (BS) optimizes discrete channel access strategies, while the low-level network in UEs learns data offloading strategies. Additionally, the computational complexity is discussed and a theoretical proof of the algorithm convergence is presented. Simulation results demonstrate the superiority of the proposed algorithm, which reduces energy consumption and data processing delay across the network.

This letter investigates the scenario of resisting intelligent reactive jammer in mobile edge computing (MEC) networks. To solve the multi-dimensional offloading allocation problem while avoiding malicious jamming attacks, a hierarchical reinforcement learning (HRL) based hybrid hidden strategy is proposed to allocate multi-dimensional resources and reduce the probability of being perceived in the offloading process. Specifically, the HRL framework is divided into two layers for asynchronous update: the upper layer with the discrete policy network taking charge of outputting the channel access strategies to resist the dynamic jamming attack in the frequency domain, and the lower layer with the continuous policy network network mainly offering hidden offloading strategies to avoid the detection of intelligent reactive jammers. By integrating the double deep Q-Network and twin delayed deep deterministic network, and exploiting their advantages in solving the hybrid strategy problems, the mobile device (MD) is able to execute effective hidden offloading strategies to reduce the computing cost as well as to significantly avoid intelligent reactive jamming in dynamic and unknown environments.

As a fundamental requirement for IoT communication systems the importance of highly reliable anti-jamming communication methods in safety use cases is growing. In this paper, we propose a novel anti-intelligent jamming scheme called opponent awareness-based anti-jamming algorithm (OA3). The user-jammer-environment interaction is formulated as a two-player simultaneous action stochastic game where participators have the ability to update their strategies. The decision-making process of each agent is modeled as a Markov decision process (MDP). Begin with the intuition “learn how the jammer learns", the opponent awareness-based iterative learning objective (OAL) of the user is presented by considering the learning awareness of the jammer to defeat the intelligent jamming. Finally, we introduce the framework including offline policy learning and online policy exploiting to implement OAL and accelerate the learning. Simulations show that the OA3 outperforms the benchmark anti-jamming strategy in terms of packet success rate.

Most of the previous research about unmanned surface vehicle (USV) control system just focuses on reaching the target but not considers berthing. Moreover, computation cost is rarely concerned. For automatic target arriving and berthing of USV in presence of multi-shape convex obstacles and disturbances, this paper designs a composite control system including trajectory optimization and trajectory tracking. In order to generate a smooth and optimal trajectory, two objectives including minimum running time and energy consumption are proposed, and their priority requirement is presented. Fuzzy satisfactory optimization is introduced to reformulate the objective functions, and the membership degree difference is used to model priority between two objectives. The obstacle avoidance constraint is developed based on the improved P-criterion. The comprehensive optimization model is established to obtain the maximum membership degree. Further, the priority optimization model is proposed to expand the feasible domain by relaxing the comprehensive optimization result. Two models are discretized by Gaussian pseudospectral method to find the optimal solution. To achieve good trajectory tracking under disturbances, two first-order active disturbance rejection control (ADRC) controllers are designed to follow the desired longitudinal velocity and course angular velocity. All the velocities of USV can converge. The simulation proves the effectiveness of the proposed co-designed system by comparing with other methods.