Xiaochen Xie

This article presents the design of a secure state estimator for systems with silent packet loss (SPL systems) and an unknown packet-loss rate (PLR). Compared with the case of observable packet loss and known PLR, silent packet loss and unknown PLR not only make the implementation of the optimal estimator (OE) infeasible, but also introduce potential security risks. This article aims to address these two issues, and the main contributions are twofold: 1) asymptotically OE: first, parameter estimators are designed to estimate the packet-loss status and the PLR, respectively. Based on this estimator, a state estimator is designed and its performance asymptotically converges to that of the OE and 2) estimator security: stable estimators are not necessarily secure in practical applications. A concept is proposed to characterize the security degree of the estimator. Then, a sufficient condition for secure estimation is established. Based on the estimator, a feedback controller is designed for an SPL system. Some simulations and experiments are presented to verify the results.

Accurate state estimation for point-foot biped robots is often compromised by significant Z-axis drift during high-frequency, dynamic motions. This paper presents KLI-Fusion, a tightly-coupled Kinematic-LiDAR-Inertial odometry framework designed to overcome this limitation. The proposed approach integrates high-frequency kinematic data, IMU measurements, and LiDAR point features within an Iterated Error State Kalman Filter (IESKF). The key innovation is an online contact-foot position enhancement module, which performs real-time terrain plane fitting to generate accurate projection constraints, thereby effectively mitigating vertical estimation drift. Extensive evaluations in both simulated and challenging real-world environments demonstrate that KLI-Fusion consistently outperforms state-of-the-art methods, reducing drift by 38% and achieving superior accuracy under strong impacts and vibrations.

Quadruped robots operating in dynamic, unstructured environments demand robust fault detection to ensure safety and autonomy. Traditional model-based approaches, while interpretable, often fail to capture the complex nonlinear dynamics of legged locomotion due to unmodeled interactions and environmental uncertainties. Data-driven methods, particularly deep learning-based techniques, have shown promise in extracting fault patterns from high-dimensional sensor data. However, existing methods often struggle with complex spatio-temporal fault patterns, limited adaptability, and catastrophic forgetting when learning new faults. To address these challenges, we propose a Dynamic Spatio-Temporal Learning framework for Fault Diagnosis in Quadruped Robots, integrating Heterogeneous Graph-Augmented Spatiotemporal Convolutional Network (HG-STCN) and Deep Reinforcement Learning. The HG-STCN explicitly models the robot's topological structure for accurate spatio-temporal fault diagnosis, while a DQN dynamically optimizes computational resource allocation to balance real-time performance and reliability. In addition, Layered Elastic Weight Consolidation) enables incremental learning, allowing the system to adapt to new faults without forgetting prior knowledge. Experimental results demonstrate superior detection accuracy, robustness, and adaptability compared to baseline methods. This work advances the reliable, long-term autonomy of quadruped robots in challenging environments.

In this article, a multiaffine approach is utilized to study the extended dissipativity performance for a class of periodic time-varying systems with constrained input and exogenous disturbances, while providing a control law to stabilize the open-loop unstable system. Specifically, a parameterized piecewise controller design based on the multiaffine property is proposed to facilitate an extended dissipativity analysis and to achieve stabilizing control for periodic time-varying systems with norm-bounded uncertainties, ensuring the closed-loop system states remain asymptotically stable under bounded energy disturbances. Moreover, the analysis of extended dissipativity can yield time-varying matrix constraints for five sets of performance indices, respectively. These constraints can also be transformed into linear matrix inequalities using the properties of multiaffine functions. The effectiveness of this approach is demonstrated through simulations of stabilization and performance indices for a helicopter model, where first-order trigonometric functions are employed to characterize bounded periodic uncertainties.

This paper studies a class of almost periodic piecewise linear systems (APPLSs), in which each cycle length (or period) is bounded yet subject to variations, making them more complex than strict periodic systems. A model reference adaptive control (MRAC) approach is developed, where the reference model is also constructed as an APPLS. Novel stability criteria are established for APPLSs, including a necessary and sufficient stability criterion for the special case of APPLSs with commuting subsystem matrices, and sufficient conditions for general cases to facilitate MRAC design. An identification technique is proposed to determine the number of subintervals in the controlled system, offering a more practical and efficient solution than existing methods. Based on that, an adaptive gain update law is provided under the formulated reference model, enabling robust adaptive control for APPLSs with variable periods and external disturbances. The proposed MRAC approach ensures state tracking of the controlled system to the desired trajectory and guarantees convergence of the tracking error. The effectiveness of the proposed approach is demonstrated through simulations based on a circuit system.

This article investigates the stability analysis and asynchronous control issues of almost periodic piecewise linear systems (APPLSs) with nonfixed periods and uncertain subinterval lengths. Unlike strictly periodic piecewise dynamic systems in previous studies, we establish new sufficient conditions using Lyapunov stability theory because there is no fixed fundamental period and each subinterval length may vary. These conditions constrain the proportion of stable and unstable subintervals to guarantee stability. Given a maximum switching delay between the controller and the subsystems, we propose a state-feedback asynchronous control approach to ensure that the closed-loop system achieves exponential stability, even when a larger proportion of subsystems are unstable. This article also considers the possibility of rapid switching between subsystems. We develop three asynchronous controller design (ACD) algorithms for cases where some subsystems are unstabilizable, all subsystems are stabilizable, and all subsystems are unstabilizable. Finally, illustrative simulations using numerical examples and a tunnel diode circuit system verify the rationality and effectiveness of the proposed approach.