Tao Xue
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Tao Xue has a total of 31 co-authors. Here are the top co-authors by number of shared publications:
- Tao Zhang
Tsinghua University
9 shared publications Bin Han
School of Mechanical Science and Engineering, Huazhong University of Science and Technology, Wuhan 430074, China
2 shared publications- Jiawei Zhao
Xian Research Institute of Hi-Tech, Xian, China
2 shared publications Meng Zhang
Department of Automation, Tsinghua University, Beijing 100084, China
2 shared publications- Ming Zhang
College of Engineering and Physical Sciences, Aston University
2 shared publications Ou Bai
Department of Electrical and Computer Engineering, Florida International University, Miami, FL 33174, USA
2 shared publications
Latest Publications
- 11 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-15
SPLIN: A Structured Plane-Based LiDAR–Inertial SLAM With Dynamic Object Removal
Accurate LiDAR-inertial SLAM in dynamic environments remains challenging due to transient moving objects and long-horizon drift accumulation. Point-only formulations offer weak structural constraints, while post-hoc dynamic filtering cannot promptly inform the frontend pose estimation and data association. To address these issues, we present SPLIN, a structured plane-based LiDAR-inertial SLAM framework that couples real-time dynamic suppression with uncertainty-consistent estimation. First, we introduce Incremental Static-referenced Dynamic Object Removal (ISDOR), a lightweight module that detects and suppresses moving objects upon their first appearance at the frame level. Second, we develop a real-time, tightly coupled Point-Plane LiDAR-Inertial Odometry (PPLIO) based on an iterated Kalman filter. PPLIO integrates ISDOR, an efficient plane extraction module, and an adaptive downsampling strategy, enabling robust operation across scenes with widely varying scales and dynamics. To mitigate long-term odometric drift, we further employ uncertainty-aware factor-graph optimization in the backend, yielding a complete, accurate, and efficient LiDAR-inertial SLAM system. Extensive experiments demonstrate the effectiveness of SPLIN: ISDOR yields, on average, a 13.47% F1 improvement (best F1 on 15/17 sequences), a 26.67% precision gain, and a 64.26% reduction in runtime over state-of-the-art baselines. Building on ISDOR, SPLIN achieves the lowest localization error on 28/34 sequences—including highly dynamic and low-feature environments—while maintaining real-time inference at 6.94-36 ms per frame. Ablation studies further validate the contributions of ISDOR and the loop closure strategy to the overall system accuracy.
- 11 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-15
SPLIN: A Structured Plane-Based LiDAR–Inertial SLAM With Dynamic Object Removal
Accurate LiDAR-inertial SLAM in dynamic environments remains challenging due to transient moving objects and long-horizon drift accumulation. Point-only formulations offer weak structural constraints, while post-hoc dynamic filtering cannot promptly inform the frontend pose estimation and data association. To address these issues, we present SPLIN, a structured plane-based LiDAR-inertial SLAM framework that couples real-time dynamic suppression with uncertainty-consistent estimation. First, we introduce Incremental Static-referenced Dynamic Object Removal (ISDOR), a lightweight module that detects and suppresses moving objects upon their first appearance at the frame level. Second, we develop a real-time, tightly coupled Point-Plane LiDAR-Inertial Odometry (PPLIO) based on an iterated Kalman filter. PPLIO integrates ISDOR, an efficient plane extraction module, and an adaptive downsampling strategy, enabling robust operation across scenes with widely varying scales and dynamics. To mitigate long-term odometric drift, we further employ uncertainty-aware factor-graph optimization in the backend, yielding a complete, accurate, and efficient LiDAR-inertial SLAM system. Extensive experiments demonstrate the effectiveness of SPLIN: ISDOR yields, on average, a 13.47% F1 improvement (best F1 on 15/17 sequences), a 26.67% precision gain, and a 64.26% reduction in runtime over state-of-the-art baselines. Building on ISDOR, SPLIN achieves the lowest localization error on 28/34 sequences—including highly dynamic and low-feature environments—while maintaining real-time inference at 6.94-36 ms per frame. Ablation studies further validate the contributions of ISDOR and the loop closure strategy to the overall system accuracy.
- 6 September 2025
- Published by World Scientific Pub Co Pte Ltd in Guidance, Navigation and Control
- Vol. 06 (02) , 153-174
Optimal Design of Network Topology for Collaborative Control of Intelligent Group Systems
The collaborative control of intelligent group systems in complex environments relies on efficient network topologies, while existing methods struggle to construct topologies with multiple optimization objectives. This paper proposes a multi-objective optimization model based on connected persistent graph, aiming to abstractly describe the topology generation problem in the formation control of intelligent group systems. By introducing the concept of connected persistent graph, we transform the traditional graph theory model into a large-scale sparse multi-objective optimization problem with strong constraints. Next, an evolutionary algorithm named SparseGR is designed for solving this problem. SparseGR innovates in population initialization, objective switching strategies, and non-dominated sorting, significantly enhancing the algorithm’s optimization ability for large-scale sparse decision variables. Finally, the proposed algorithm is verified in four typical formations and large-scale random formations. Experimental results show that SparseGR outperforms existing algorithms in terms of the topology’s robustness against damage, formation maintenance capability, and rapid response capability. In addition, the algorithm demonstrates advantages in convergence, effectiveness, and timeliness, with improvements of at least 9.86%, 19.44%, and 6.85%, respectively.
- 6 September 2025
- Published by World Scientific Pub Co Pte Ltd in Guidance, Navigation and Control
- Vol. 06 (02) , 153-174
Optimal Design of Network Topology for Collaborative Control of Intelligent Group Systems
The collaborative control of intelligent group systems in complex environments relies on efficient network topologies, while existing methods struggle to construct topologies with multiple optimization objectives. This paper proposes a multi-objective optimization model based on connected persistent graph, aiming to abstractly describe the topology generation problem in the formation control of intelligent group systems. By introducing the concept of connected persistent graph, we transform the traditional graph theory model into a large-scale sparse multi-objective optimization problem with strong constraints. Next, an evolutionary algorithm named SparseGR is designed for solving this problem. SparseGR innovates in population initialization, objective switching strategies, and non-dominated sorting, significantly enhancing the algorithm’s optimization ability for large-scale sparse decision variables. Finally, the proposed algorithm is verified in four typical formations and large-scale random formations. Experimental results show that SparseGR outperforms existing algorithms in terms of the topology’s robustness against damage, formation maintenance capability, and rapid response capability. In addition, the algorithm demonstrates advantages in convergence, effectiveness, and timeliness, with improvements of at least 9.86%, 19.44%, and 6.85%, respectively.
- 10 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automation Science and Engineering
- Vol. 22 (15455955) , 16572-16590
Confidence-Aware Robust Dynamical Distance Constrained Reinforcement Learning for Social Robot Navigation
Navigating in a crowded social environment without collisions or freezing is a crucial and challenging task. Recent studies have demonstrated considerable success using Deep Reinforcement Learning for social robot navigation. However, the lack of safety guarantees remains a prevalent problem among these methods. To address this limitation, we present a comprehensive safety framework that builds upon a novel multi-agent environment single-agent decision-making formulation and the strengths of Safe Reinforcement Learning (Safe RL) methods. First, considering the uncertainty of pedestrians and the inaccuracy of trajectory predictions, we explicitly embed confidence-weighted uncertainty and risk measurement in the interaction graph network to induce adaptive policies. Next, we replace the state-wise robust-form unshaped cost with a cumulative robust dynamical distance constraint, which establishes a stronger and essential safety objective while ensuring the invariance of the optimal policy. Then we use reachability-based Safe RL to ensure state-wise safety assurance of the feasible part and the risk minimization of the infeasible part. Simulation and virtual-reality evaluation results demonstrate that our proposed method outperforms the state-of-the-art methods in various metrics, such as success rate and collision energy, achieving the best trade-off between safety and efficiency. Our proposed method yields a policy that exhibits desired collision avoidance behavior in both simulation and real-world experiments.
- 10 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automation Science and Engineering
- Vol. 22 (15455955) , 16572-16590
Confidence-Aware Robust Dynamical Distance Constrained Reinforcement Learning for Social Robot Navigation
Navigating in a crowded social environment without collisions or freezing is a crucial and challenging task. Recent studies have demonstrated considerable success using Deep Reinforcement Learning for social robot navigation. However, the lack of safety guarantees remains a prevalent problem among these methods. To address this limitation, we present a comprehensive safety framework that builds upon a novel multi-agent environment single-agent decision-making formulation and the strengths of Safe Reinforcement Learning (Safe RL) methods. First, considering the uncertainty of pedestrians and the inaccuracy of trajectory predictions, we explicitly embed confidence-weighted uncertainty and risk measurement in the interaction graph network to induce adaptive policies. Next, we replace the state-wise robust-form unshaped cost with a cumulative robust dynamical distance constraint, which establishes a stronger and essential safety objective while ensuring the invariance of the optimal policy. Then we use reachability-based Safe RL to ensure state-wise safety assurance of the feasible part and the risk minimization of the infeasible part. Simulation and virtual-reality evaluation results demonstrate that our proposed method outperforms the state-of-the-art methods in various metrics, such as success rate and collision energy, achieving the best trade-off between safety and efficiency. Our proposed method yields a policy that exhibits desired collision avoidance behavior in both simulation and real-world experiments.
- 20 November 2023
- Published by Cambridge University Press (CUP) in Robotica
- Vol. 42 (2) , 367-388
Interpretable long-term gait trajectory prediction based on Interpretable-Concatenation former
Human gait trajectory prediction is a long-standing research topic in human–machine interaction. However, there are two shortcomings in the current gait trajectory prediction technology. The first shortcoming is that the neural network model of gait prediction only predicts dozens of future time frames of gait trajectory. The second shortcoming is that the gait prediction neural network model is uninterpretable. We propose the Interpretable-Concatenation former (IC-former) model, which can predict long-term gait trajectories and explain the prediction results by quantifying the importance of data at different positions in the input sequence. Experiments prove that the IC-former model we proposed not only makes a breakthrough in prediction accuracy but also successfully explains the data basis of the prediction.
- 20 November 2023
- Published by Cambridge University Press (CUP) in Robotica
- Vol. 42 (2) , 367-388
Interpretable long-term gait trajectory prediction based on Interpretable-Concatenation former
Human gait trajectory prediction is a long-standing research topic in human–machine interaction. However, there are two shortcomings in the current gait trajectory prediction technology. The first shortcoming is that the neural network model of gait prediction only predicts dozens of future time frames of gait trajectory. The second shortcoming is that the gait prediction neural network model is uninterpretable. We propose the Interpretable-Concatenation former (IC-former) model, which can predict long-term gait trajectories and explain the prediction results by quantifying the importance of data at different positions in the input sequence. Experiments prove that the IC-former model we proposed not only makes a breakthrough in prediction accuracy but also successfully explains the data basis of the prediction.
- 1 September 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Hybrid Gait Recognition Method Based on Phase Trajectory
Gait recognition is critical to the activity monitoring, health management, assistance control of prostheses and exoskeletons, etc. This study aims to improve the gait classification performance on daily hybrid locomotions. We found the hip angle phase trajectories present significantly gait-dependent patterns, whereas the phase patterns show repeated limit cycles for periodic gaits while half-cycles or dots for aperiodic gaits. By converting the gait recognition issue into an image classification problem, we propose to use a convolution neural network (CNN) to learn the gait-dependent phase pattern images. Besides, to enhance the gait transition stability, we further integrate the prior state transition probability with category likelihood via dynamic Bayesian network. The proposed method has been experimented with 6 healthy subjects on standing, sitting, stand-to-sit, sit-to-stand, walking, running, stair ascending, and stair descending gaits. The overall leave-one-out cross-validation accuracy in continuous time is 96.15%.
- 1 September 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Hybrid Gait Recognition Method Based on Phase Trajectory
Gait recognition is critical to the activity monitoring, health management, assistance control of prostheses and exoskeletons, etc. This study aims to improve the gait classification performance on daily hybrid locomotions. We found the hip angle phase trajectories present significantly gait-dependent patterns, whereas the phase patterns show repeated limit cycles for periodic gaits while half-cycles or dots for aperiodic gaits. By converting the gait recognition issue into an image classification problem, we propose to use a convolution neural network (CNN) to learn the gait-dependent phase pattern images. Besides, to enhance the gait transition stability, we further integrate the prior state transition probability with category likelihood via dynamic Bayesian network. The proposed method has been experimented with 6 healthy subjects on standing, sitting, stand-to-sit, sit-to-stand, walking, running, stair ascending, and stair descending gaits. The overall leave-one-out cross-validation accuracy in continuous time is 96.15%.
- 13 August 2020
- Published by Cambridge University Press (CUP) in Robotica
- Vol. 39 (2) , 291-316
Continuous Finite-Time Torque Control for Flexible Assistance Exoskeleton with Delay Variation Input
SUMMARY: Accurate torque control is a critical issue in the compliant human–robot interaction scenario, which is, however, challenging due to the ever-changing human intentions, input delay, and various disturbances. Even worse, the performances of existing control strategies are limited on account of the compromise between precision and stability. To this end, this paper presents a novel high-performance torque control scheme without compromise. In this scheme, a new nonlinear disturbance observer incorporated with equivalent control concept is proposed, where the faster convergence and stronger anti-noise capability can be obtained simultaneously. Meanwhile, a continuous fractional power control law is designed with an iteration method to address the matched/unmatched disturbance rejection and global finite-time convergence. Moreover, the finite-time stability proof and prescribed control performance are guaranteed using constructed Lyapunov function with adding power integrator technique. Both the simulation and experiments demonstrate enhanced control accuracy, faster convergence rate, perfect disturbance rejection capability, and stronger robustness of the proposed control scheme. Furthermore, the evaluated assistance effects present improved gait patterns and reduced muscle efforts during walking and upstair activity.
- 13 August 2020
- Published by Cambridge University Press (CUP) in Robotica
- Vol. 39 (2) , 291-316
Continuous Finite-Time Torque Control for Flexible Assistance Exoskeleton with Delay Variation Input
SUMMARY: Accurate torque control is a critical issue in the compliant human–robot interaction scenario, which is, however, challenging due to the ever-changing human intentions, input delay, and various disturbances. Even worse, the performances of existing control strategies are limited on account of the compromise between precision and stability. To this end, this paper presents a novel high-performance torque control scheme without compromise. In this scheme, a new nonlinear disturbance observer incorporated with equivalent control concept is proposed, where the faster convergence and stronger anti-noise capability can be obtained simultaneously. Meanwhile, a continuous fractional power control law is designed with an iteration method to address the matched/unmatched disturbance rejection and global finite-time convergence. Moreover, the finite-time stability proof and prescribed control performance are guaranteed using constructed Lyapunov function with adding power integrator technique. Both the simulation and experiments demonstrate enhanced control accuracy, faster convergence rate, perfect disturbance rejection capability, and stronger robustness of the proposed control scheme. Furthermore, the evaluated assistance effects present improved gait patterns and reduced muscle efforts during walking and upstair activity.
- 21 May 2020
- Published by Zhejiang University Press in Frontiers of Information Technology & Electronic Engineering
- Vol. 21 (5) , 705-722
Fixed-time constrained acceleration reconstruction scheme for robotic exoskeleton via neural networks
Accurate acceleration acquisition is a critical issue in the robotic exoskeleton system, but it is difficult to directly obtain the acceleration via the ex
- 21 May 2020
- Published by Zhejiang University Press in Frontiers of Information Technology & Electronic Engineering
- Vol. 21 (5) , 705-722
Fixed-time constrained acceleration reconstruction scheme for robotic exoskeleton via neural networks
Accurate acceleration acquisition is a critical issue in the robotic exoskeleton system, but it is difficult to directly obtain the acceleration via the ex
- 3 July 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 4 (4) , 3318-3323
Adaptive Oscillator-Based Robust Control for Flexible Hip Assistive Exoskeleton
This study aims to develop a wearable robotic exoskeleton to provide walking assistance for elderly and disabled people with weakened skeletal muscles. Our developed hip exoskeleton is powered by two electrical series elastic actuators covering both hip extension/flexion joints. For exoskeleton control, motion intention perception, delivering appropriate force and flexible human-exoskeleton interaction still remains challenging. To this end, this paper proposes an improved gait synchronizer driven by the only hip angular velocity to synchronize gait phase and further estimate gait features. To adapt to the ever-changing gait, the commanded torque from the nominal biological moment is modulated with estimated walking states and speed. The sEMG acquisition experimental results from 6 healthy subjects demonstrate its effectiveness on movement assistance, i.e., reducing muscle activation efforts significantly for 4Km/h walking and upstairs.
- 3 July 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 4 (4) , 3318-3323
Adaptive Oscillator-Based Robust Control for Flexible Hip Assistive Exoskeleton
This study aims to develop a wearable robotic exoskeleton to provide walking assistance for elderly and disabled people with weakened skeletal muscles. Our developed hip exoskeleton is powered by two electrical series elastic actuators covering both hip extension/flexion joints. For exoskeleton control, motion intention perception, delivering appropriate force and flexible human-exoskeleton interaction still remains challenging. To this end, this paper proposes an improved gait synchronizer driven by the only hip angular velocity to synchronize gait phase and further estimate gait features. To adapt to the ever-changing gait, the commanded torque from the nominal biological moment is modulated with estimated walking states and speed. The sEMG acquisition experimental results from 6 healthy subjects demonstrate its effectiveness on movement assistance, i.e., reducing muscle activation efforts significantly for 4Km/h walking and upstairs.
- 1 January 2018
- Published by AIP Publishing in Review of Scientific Instruments
- Vol. 89 (1) , 013904
Graphical method for analyzing wide-angle x-ray diffraction
Wide-angle X-ray diffraction on large-scale laser facility is a well-established experimental method, which is used to study the shock response of single crystal materials by recording X-rays diffracted from numerous lattice planes. We present a three-dimensional graphical method for extracting physical understanding from the raw diffraction data in shocked experiments. This method advances beyond the previous iterative process by turning abstract diffraction theories in shock physics into mathematic issues, providing three-dimensional visualization and quick extraction of data characteristics. The capability and versatility of the method are exhibited by identifying lattice planes for single crystal samples with different orientations and quantitatively measuring the lattice compression and rotation under dynamic loading.
- 1 January 2018
- Published by AIP Publishing in Review of Scientific Instruments
- Vol. 89 (1) , 013904
Graphical method for analyzing wide-angle x-ray diffraction
Wide-angle X-ray diffraction on large-scale laser facility is a well-established experimental method, which is used to study the shock response of single crystal materials by recording X-rays diffracted from numerous lattice planes. We present a three-dimensional graphical method for extracting physical understanding from the raw diffraction data in shocked experiments. This method advances beyond the previous iterative process by turning abstract diffraction theories in shock physics into mathematic issues, providing three-dimensional visualization and quick extraction of data characteristics. The capability and versatility of the method are exhibited by identifying lattice planes for single crystal samples with different orientations and quantitatively measuring the lattice compression and rotation under dynamic loading.