ZMZhongjing Ma

Zhongjing Ma

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Zhongjing Ma has a total of 68 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • DRIVE-Nav: Directional Reasoning, Inspection, and Verification for Efficient Open-Vocabulary Navigation

    by Maoguo Gao, Zejun Zhu, Zhiming Sun, Zhengwei Ma, Longze Yuan, Zhongjing Ma, Zhigang Gao, Jinhui Zhang, Suli Zou

  • GVF-MPC Depth Control Framework Motivated by Restoring-Moment Mechanism Analysis for Bionic Robotic Fish

    by Jiarong Han, Xiangqing Yuan, Bo Yin, Yu Liu, Suli Zou, Zhihong Deng, Zhongjing Ma

    Longitudinal oscillations are commonly observed in depth control of low-speed underwater platforms such as bio-inspired robotic fish. To investigate this issue, this letter conducts a quantitative analysis of the vertical-plane dynamics. It reveals the speed-dependent coupling between the depth and pitch channels as the swimming speed decreases, and clarifies the underdamped mechanism induced by restoring-moment dominance. A critical speed condition for the onset of low-speed oscillations is also derived. Based on the dynamic analysis, a hierarchical GVF–MPC depth control framework is proposed, which structurally decouples geometric guidance from dynamic execution. The guidance layer employs a guiding vector field (GVF) to generate naturally smooth attitude references from geometric information, with reduced sensitivity to model degradation and pitch disturbances. The inner loop applies model predictive control (MPC) to track the guidance commands while explicitly accounting for dynamic constraints, balancing convergence speed and control-input smoothness to improve stability and safety under low-speed conditions. Simulation and experiments are conducted to evaluate the effectiveness and robustness of the proposed method. Simulations consider sinusoidal and V-shaped depth path and compare the proposed approach with LOS, PID, and SMC based controllers. Experiments are performed in still water and under simulated Sea-State 1 disturbances to assess steady state accuracy, disturbance recovery, and overall robustness.


  • Vision-Based Obstacle Avoidance and Formation Control for Underwater Robotic Fish

    by Jiarong Han, Rui Huang, Xiangqing Yuan, Yu Liu, Bo Yin, Suli Zou, Zhongjing Ma

    This letter aims to enhance the autonomy of bionic robotic fish formation in executing underwater tasks by integrating visual recognition and control systems. Firstly, we design an adaptive image enhancement (AIE) module that integrates a hyperparameter neural network (HPNN) into the YOLOv8 framework, which improves recognition performance under low-light and high-interference underwater conditions, enhancing the obstacle perception capability of underwater robots. Secondly, considering the underactuated and nonlinear hydrodynamic characteristics of the robotic fish system, a formation heading and speed controller with constrained control freedom is designed. This involves establishing a side-slip dynamics model for the robotic fish, analyzing its nonlinear hydrodynamics, and proving the Lyapunov stability of the controller. Finally, the synergistic efficacy of the visual and control systems is validated through a series of experiments, including target tracking, target frame traversal, formation maintenance and reconstruction, and formation obstacle avoidance. These experiments demonstrate that the collaboration of the proposed perception and control modules significantly enhances the capability of the robotic fish formation to autonomously undertake complex underwater tasks.


  • Game-Based Social Learning Particle Swarm Optimizer for Inverse Kinematics of Robotic Arms

    by Kexin Hu, Zhongjing Ma, Suli Zou, Jian Li, Jinhui Zhang

    Robot inverse kinematics is the foundation of robotics and an indispensable part of robot development and applications. The inverse kinematics of a robotic arm involves its specific configuration, the non-convex coupling relationships between joints, and the presence of multiple solutions, among other factors. Existing works usually focus on the robotic arm as a whole, emphasizing the pose accuracy of the robot's end-effector, without considering the individuality of each joint. In this letter, the inverse kinematics problem of robotic arms is formulated as a non-convex optimization problem, with the goal of optimizing the contribution of each joint. Game theory is rigorously introduced to convert the optimization problem into a joint game Nash equilibrium (NE) problem. A generalized Lagrange method is used to handle coupling constraints by introducing additional penalty terms into the objective function. We propose a game-based social learning particle swarm optimization (GSLPSO) algorithm that combines NE strategy and social learning mechanism to enhance the problem-solving process. The proposed GSLPSO is compared with existing relevant algorithms and verified on a real-world robotic arm AUBO i16, which proves the superiority and practicality of the proposed algorithm.


  • Data-Assisted Dynamic Modeling of Bionic Robotic Fish and Its Precise Speed Control

    by Jiarong Han, Shun Huang, Yingyu Yao, Zhongjing Ma, Yu Liu, Suli Zou, Bo Yin

    Dynamic modeling is essential for comprehending physical mechanisms and devising control strategies in bionic robot research. This letter introduces a novel dynamic modeling method that combines Lagrangian dynamics and data-assisted techniques for robotic fish with bionic morphology, multi-joint structures, and a flexible caudal fin. Firstly, a nonlinear continuous hydrodynamic model has been developed using extensive data derived from computational fluid dynamics (CFD), thereby capturing the high-fidelity locomotion of robotic fish. Secondly, based on mathematical derivation, a stability analysis method and controller design approach for biomimetic systems with periodic behaviors have been proposed. Furthermore, to demonstrate the model's efficacy, we designed a model reference adaptive controller for speed control. Both simulation and experimental results validate the model's accuracy, effectiveness, and potential for improving control consistency in tracking time-varying speeds of robotic fish.


  • Distributed Momentum-Based Multiagent Optimization With Different Constraint Sets

    by Xu Zhou, Zhongjing Ma, Suli Zou, Kostas Margellos

    In this article, we consider a class of consensus optimization problems over a time-varying communication network wherein each agent can only interact with its neighbors. The target is to minimize the summation of all local and possibly nonsmooth objectives in the presence of different constraint sets per agent. To achieve this goal, we propose a novel distributed heavy-ball algorithm that combines the subgradient tracking technique with a momentum term related to history information. This algorithm promotes the distributed application of existing centralized accelerated momentum methods, especially for constrained nonsmooth problems. Under certain assumptions and conditions on the step-size and momentum coefficient, the convergence and optimality of the proposed algorithm can be guaranteed through a rigorous theoretical analysis, and a convergence rate of $\mathcal {O}(\text {ln}k/ \sqrt{k})$ in objective value is also established. Simulations on an $\ell _{1}$-regularized logistic-regression problem show that the proposed algorithm can achieve faster convergence than existing related distributed algorithms, while a case study involving a building energy management problem further demonstrates its efficacy.


  • Attack Detection and Mitigation Scheme of Load Frequency Control Systems Against False Data Injection Attacks

    by Xin Zhao, Zhongjing Ma, Xingyu Shi, Suli Zou

    Load frequency control (LFC) of modern power systems depends on open communication networks to transmit measurement/control signals, which may make the LFC scheme vulnerable to cyberattacks. Therefore, it is necessary to ensure the resilience of LFC systems against cyberattacks so as to detect the attack and maintain the stable operation of the system during the attack. This article aims to investigate the attack detection and mitigation scheme of LFC systems against false data injection attacks. A model-based unknown input observer (UIO) is designed to simultaneously estimate the system state and attack signal based on load prediction, where the attack estimation is utilized to achieve real-time attack detection. Moreover, we present a sufficient and necessary condition for the existence of the UIO and a design method of the UIO based on linear matrix inequalities. Based on the attack estimation, a resilient LFC scheme is proposed to mitigate the attack impact and restore the frequency and tie-line power within an accepted range during the attack. Finally, case studies are undertaken by the IEEE 9- and 39-bus test systems to show the effectiveness of the proposed scheme.


  • Joint Energy and Carbon Trading for Multi-Microgrid System Based on Multi-Agent Deep Reinforcement Learning

    by Yanting Zhou, Zhongjing Ma, Tianyu Wang, Jinhui Zhang, Xingyu Shi, Suli Zou

    Carbon trading has emerged as an effective way to promote the renewable generation and sustainable energy development. Since carbon emissions are closely coupled to energy system, it is a challenge to design a market mechanism for joint energy and carbon trading to achieve better strategies. In this study, the energy management problem with a specific focus on joint trading in multi-microgrid system is investigated by utilizing a multi-agent deep reinforcement learning approach. Initially, a joint energy and carbon trading market is established and the dispatch optimization problem is formulated as a Markov decision process without modeling uncertainties accurately. This mechanism enables direct one-to-one energy transactions among all areas, avoiding the market clearing in traditional multi-party local energy trading markets. To enhance the learning efficiency and maintain agent privacy, an enhanced multi-agent proximal policy optimization (MAPPO) algorithm that incorporates a parameter sharing mechanism is introduced. Moreover, the recurrent neural networks (RNN) structure is leveraged to perform feature encoding for individual agents, which improves the overall feature extraction capability. Through comprehensive experiments involving various algorithms, the proposed approach reduce operating costs 14.86 $\%$ and carbon emissions 19.04 $\%$ compared with traditional MAPPO, which validates the effectiveness and performance benefits.


  • Open Access

    Winnie: Task-Oriented Dialog System with Structure-Aware Contrastive Learning and Enhanced Policy Planning

    by Kaizhi Gao, Tianyu Wang, Zhongjing Ma, Suli Zou

    Pre-trained encoder-decoder models are widely applied in Task-Oriented Dialog (TOD) systems on the session level, mainly focusing on modeling the dialog semantic information. Dialogs imply structural information indicating the interaction among user utterances, belief states, database search results, system acts and responses, which is also crucial for TOD systems. In addition, for the system acts, additional pre-training and datasets are considered to improve their accuracies, undoubtedly introducing a burden. Therefore, a novel end-to-end TOD system named Winnie is proposed in this paper to improve the TOD performance. First, to make full use of the intrinsic structural information, supervised contrastive learning is adopted to narrow the gap in the representation space between text representations of the same category and enlarge the overall continuous representation margin between text representations of different categories in dialog context. Then, a system act classification task is introduced for policy optimization during fine-tuning. Empirical results show that Winnie substantially improves the performance of the TOD system. By introducing the supervised contrastive and system act classification losses, Winnie achieves state-of-the-art results on benchmark datasets, including MultiWOZ2.2, In-Car, and Camrest676. Their end-to-end combined scores are improved by 3.2, 1.9, and 1.1 points, respectively.