XSXiongtao Shi

Xiongtao Shi

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

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Xiongtao Shi has a total of 20 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Future-Trend-Aware Filter-Based PD-MRAC Method for Quadrotors With Unknown Strong Disturbances

    by Yanhua Yang, Chenxin Yu, Xiongtao Shi, Changchun Hua, James Lam, Youmin Gong, Jie Mei

    Robust flight in complex and windy environments is critical for both single and multiple quadrotors. Existing methods either learn disturbance model at high computational cost or use error-based adaptive control with a speed-stability trade-off that makes tuning difficult. To address these issues, this paper proposes a future-trend-aware filter-based PD-MRAC (Proportional-Derivative Model Reference Adaptive Control) for single quadrotor and a distributed PD-MRAC for multiple quadrotor formation. By embedding a trend-aware derivative term in the adaptive update laws, the controller obtains anticipatory information about the error evolution, enabling rapid adaptation while mitigating oscillations. For more disturbance-sensitive multi-quadrotors, we design a robust distributed protocol under a directed graph, improving resilience to disturbances. The approach maintains low computational cost and supports fast adaptive updates. Extensive simulations and real-world experiments validate improvement. For single quadrotor, RMSE reduced by around 57% versus the baselines and by around 12% versus the DJI Mavic 2. For multi-quadrotors, formation results show enhanced robustness in simulation and effective real-world indoor/outdoor experiments under strong winds. Our project page is at https://xiongtao-shi.github.io/PD-MRAC/.

  • Fully Distributed Event-Triggered Control for Containment of Fully Heterogeneous MASs Using Filter-Based RL

    by Yanhua Yang, Jie Mei, Xiongtao Shi, Guangfu Ma

    The complexity of multi-agent systems (MASs) poses challenges in containment control, especially with fully heterogeneous agents influenced by multiple leaders. Traditional methods rely on continuous communication, global information, or homogeneous dynamics, limiting real-world applicability. In this paper, a model-free reinforcement learning (RL) based fully distributed event-triggered control (ETC) framework is proposed. First, an integrated ETC framework is designed for directed graphs, allowing agents to observe multiple leaders simultaneously. To further reduce the communication burden, an improved fully distributed ETM with $\sigma$-modification is introduced, eliminating global information dependency and enhancing communication efficiency. Based on the observed information, for unknown heterogeneous dynamics, a filter-based model-free RL algorithm solves a fully distributed normalized level of influence (NLI) based augmented algebraic Riccati equation (AARE), enabling online learning from input-output data without system modeling while reducing computational complexity. Simulations validate the framework's effectiveness, demonstrating improved control scalability, communication efficiency, and adaptability due to its fully model-free nature.


  • Fully Distributed Consensus of Multiple Euler–Lagrange Systems With Time-Varying Asymmetric Full-State Constraints

    by Yanhua Yang, Jie Mei, Xiongtao Shi, Guangfu Ma

    In this article, the consensus problem of multiple Euler–Lagrange (EL) systems with time-varying asymmetric full-state constraints under a directed graph is investigated in a fully distributed way, where the global information dependence is removed. First, to prevent the violation of the full-state constraints of each EL system, a nonlinear state-dependent transformation is adopted for both the leader and followers, where the constraints will not be violated as long as the transformed states remain finite. Then, a constrained nonlinear fully distributed observer is constructed using only neighbors' states, and an adaptive gain is designed to eliminate the dependence of the global information. The theoretical analysis is completed via constructing a novel integration-based Lyapunov function using a graph-based diagonal matrix, ensuring the achievement of consensus. Using the observed state, an adaptive backstepping state tracking control protocol is developed such that the position of each EL system can track the leader within a tolerable bounded region. Thus, the leader-following consensus of multiple EL systems with time-varying asymmetric full-state constraints is achieved. Simulation results are illustrated to show the feasibility and effectiveness of the proposed control scheme.


  • Filter-Based Fully Distributed Output Regulation of Heterogeneous Learning Agents

    by Xiongtao Shi, Yanjie Li, Chenglong Du, Chaoyang Chen, Changchun Hua, Weihua Gui

    This paper proposes a novel filter-based model-free reinforcement learning (RL) event-triggered control (ETC) method for the fully distributed robust leaderless cooperative output regulation (COR) of unknown heterogeneous multi-agent systems (MASs) with external disturbances over directed graphs. First, the fully distributed event-triggered observers are designed to generate an autonomous system for the robust leaderless COR, in which the frequency of signal transmission and computational burden are significantly reduced, and the Zeno behavior is strictly ruled out. Then, a filter-based model-free RL algorithm without integration operation is developed to obtain the solution of the internal model-based augmented algebraic Riccati equation (AARE) and to release the requirement of recording complete and continuous data. Moreover, with some adaptive parameters, the robust leaderless COR is solved in a fully distributed manner without involving any global information of directed MASs. Finally, simulation results on RLC circuits are illustrated to show the feasibility and effectiveness of the proposed control scheme.


  • Fully Distributed Event-Triggered Control of Nonlinear Multiagent Systems Under Directed Graphs: A Model-Free DRL Approach

    by Xiongtao Shi, Yanjie Li, Chenglong Du, Yang Shi, Chunhua Yang, Weihua Gui

    This article addresses the consensus problem of a class of unknown nonlinear multi-agent systems (MASs) under directed graphs via a novel model-free deep reinforcement learning (DRL) based fully distributed event-triggered control (ETC) method. First, the DRL-based feedback linearization approach is developed to learn an approximated linearized control protocol in a model-free manner. Then, a novel adaptive event-triggered mechanism (ETM) is proposed to save more communication resources and reduce the computational burden among agents, and the Zeno behavior is ruled out strictly. The control protocol proposed in this paper does not involve global information, thus it can be implemented in a fully distributed manner. Furthermore, a new Lyapunov function is constructed using a graph-based diagonal matrix to achieve the consensus of MASs under directed graphs. Generally, distinct from the existing results, the proposed model-free DRL-based fully distributed ETC protocol has the following features: 1) only using the intermittent local information, 2) not requiring the model information and global graph information, and 3) applicable to the more general directed graph. Finally, simulation results are illustrated to show the feasibility and effectiveness of the proposed control scheme.


  • Distributional Policy Gradient With Distributional Value Function

    by Qi Liu, Yanjie Li, Xiongtao Shi, Ke Lin, Yuecheng Liu, Yunjiang Lou

    In this article, we propose a distributional policy-gradient method based on distributional reinforcement learning (RL) and policy gradient. Conventional RL algorithms typically estimate the expectation of return, given a state–action pair. Furthermore, distributional RL algorithms consider the return as a random variable and estimate the return distribution that can characterize the probability of different returns resulted by environmental uncertainties. Thus, the return distribution provides more valuable information than its expectation, leading to superior policies in general. Although distributional RL has been investigated widely in value-based RL methods, very few policy-gradient methods take advantage of distributional RL. To bridge this research gap, we propose a distributional policy-gradient method by introducing a distributional value function to the policy gradient (DVDPG). We estimate the distribution of policy gradient instead of the expectation estimated in conventional policy-gradient RL methods. Furthermore, we propose two policy-gradient value sampling mechanisms to do policy improvement. First, we propose a distribution-probability-sampling method that samples the policy-gradient value according to the quantile probability of return distribution. Second, a uniform sample mechanism is proposed. With our sample mechanisms, the proposed distributional policy-gradient method enhances the stochasticity of the policy gradient, improving the exploration efficiency and benefiting to avoid falling into local optimal solutions. In sparse-reward tasks, the distribution-probability-sampling method outperforms the uniform sample mechanism. In dense-reward tasks, the two sample mechanisms perform similarly. Moreover, we show that the conventional policy-gradient method is a special case of the proposed method. Experimental results on various sparse-reward and dense-reward OpenAI-gym tasks illustrate the efficiency of the proposed method, outperforming baselines in almost environments.


  • A Time-Aggregated Model-Free RL Algorithm for Optimal Containment Control of MASs

    by Xiongtao Shi, Yanjie Li, Chenglong Du, Weihua Gui

    In this paper, the optimal containment control problem for a class of unknown nonlinear multi-agent systems (MASs) is studied via a time-aggregated (TA) model-free reinforcement learning (RL) algorithm. First, based on the idea of TA, the control policy is updated only when the system visits a finite subset of the state space. Thus, the control is event-triggered and not time-triggered. On this basis, a model-free TA-based value iteration (TA-VI) algorithm is proposed to learn the optimal control protocol. Since the finite important states are considered and the control is event-triggered, this algorithm requires fewer updating times and fewer computation than the conventional optimal containment control. Moreover, the TA-VI algorithm eliminates requirements on the function approximator and state discretization, which allows a strict convergence analysis via the mathematical induction method. Finally, simulation results are given to show the feasibility and superiority of the proposed algorithm.


  • Optimal Lateral Path-Tracking Control of Vehicles With Partial Unknown Dynamics via DPG-Based Reinforcement Learning Methods

    by Xiongtao Shi, Yanjie Li, Wenxiao Hu, Chenglong Du, Chaoyang Chen, Weihua Gui

    This article focuses on the optimal lateral path-tracking control problem of vehicles with unknown drift dynamics in a model-free manner through two novel deterministic policy gradient (DPG) based reinforcement learning (RL) methods. First, due to the difficulty of modeling the precise dynamics of vehicles, a policy gradient (PG) is derived to learn the optimal control gain by minimizing a predefined infinite-horizon performance index, where the knowledge of the system drift dynamics of vehicles is no longer needed. Then, to further remove the limitation of the initial admissibility of the control policy, a two-stage DPG-based RL optimal control algorithm is proposed, in which a novel finite-horizon performance index is employed in the pre-learning stage such that the control gain does not require to be initially admissible. It should be pointed out that the derived PGs in the two algorithms are based on an explicit form only using a single sampling data for each calculation rather than an estimated form via randomly perturbing feedback gains, which reduces the sampling and computational complexity of the algorithms. Finally, the simulations of the lateral path-tracking control of vehicles have verified the effectiveness and superiority of the proposed DPG-based RL algorithms compared with existing methods.