Xin Liu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Xin Liu has a total of 35 co-authors. Here are the top co-authors by number of shared publications:
- Chenkun Qi
Shanghai Jiao Tong University
5 shared publications - Feng Gao
Ocean University of China, Qingdao, China
4 shared publications - Jianfeng Lin
State Key Laboratory of Mechanical System and Vibration, School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai, China
2 shared publications Nuo Xu
Ecole dingenieurs Paris, Shanghai Jiao Tong University, Shanghai, China
2 shared publicationsYule Mo
State Key Laboratory of Mechanical System and Vibration, School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai, China
2 shared publications- Aoqian Zhang
National University of Singapore
1 shared publication
Latest Publications
- 1 January 2027
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 12 (1) , 17-24
E2DRL: End-to-End Differentiable Reinforcement Learning for 3D Obstacle Avoidance in Quadruped Robots
- 1 January 2027
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 12 (1) , 17-24
E2DRL: End-to-End Differentiable Reinforcement Learning for 3D Obstacle Avoidance in Quadruped Robots
- 11 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11 (2) , 1666-1673
PPL: Point Cloud Supervised Proprioceptive Locomotion Reinforcement Learning for Legged Robots in Crawl Spaces
Legged locomotion in constrained spaces (called crawl spaces) is challenging. In crawl spaces, current proprioceptive locomotion learning methods are difficult to achieve traverse because only ground features are inferred. In this study, a point cloud supervised RL framework for proprioceptive locomotion in crawl spaces is proposed. A state estimation network is designed to estimate the robot’s collision states as well as ground and spatial features for locomotion. A point cloud feature extraction method is proposed to supervise the state estimation network. The method uses representation of the point cloud in polar coordinate frame and MLPs for efficient feature extraction. Experiments demonstrate that, compared with existing methods, our method exhibits faster iteration time in the training and more agile locomotion in crawl spaces. This study enhances the ability of legged robots to traverse constrained spaces without requiring exteroceptive sensors.
- 11 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11 (2) , 1666-1673
PPL: Point Cloud Supervised Proprioceptive Locomotion Reinforcement Learning for Legged Robots in Crawl Spaces
Legged locomotion in constrained spaces (called crawl spaces) is challenging. In crawl spaces, current proprioceptive locomotion learning methods are difficult to achieve traverse because only ground features are inferred. In this study, a point cloud supervised RL framework for proprioceptive locomotion in crawl spaces is proposed. A state estimation network is designed to estimate the robot’s collision states as well as ground and spatial features for locomotion. A point cloud feature extraction method is proposed to supervise the state estimation network. The method uses representation of the point cloud in polar coordinate frame and MLPs for efficient feature extraction. Experiments demonstrate that, compared with existing methods, our method exhibits faster iteration time in the training and more agile locomotion in crawl spaces. This study enhances the ability of legged robots to traverse constrained spaces without requiring exteroceptive sensors.
- 13 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11 (1) , 81-88
MLM: Learning Multi-Task Loco-Manipulation Whole-Body Control for Quadruped Robot With Arm
Whole-body loco-manipulation for quadruped robots with arms remains a challenging problem, particularly in achieving multi-task control. To address this, we propose MLM, a reinforcement learning framework driven by both real-world and simulation data. It enables a six-DoF robotic arm–equipped quadruped robot to perform whole-body loco-manipulation for multiple tasks autonomously or under human teleoperation. To address the problem of balancing multiple tasks during the learning of loco-manipulation, we introduce a trajectory library with an adaptive, curriculum-based sampling mechanism. This approach allows the policy to efficiently leverage real-world collected trajectories for learning multi-task loco-manipulation. To address deployment scenarios with only historical observations and to enhance the performance of policy execution across tasks with different spatial ranges, we propose a Trajectory-Velocity Prediction policy network. It predicts unobservable future trajectories and velocities. By leveraging extensive simulation data and curriculum-based rewards, our controller achieves whole-body behaviors in simulation and zero-shot transfer to real-world deployment. Ablation studies in simulation verify the necessity and effectiveness of our approach, while real-world experiments on a Go2 robot with an Airbot robotic arm demonstrate the policy's good performance in multi-task execution.
- 13 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11 (1) , 81-88
MLM: Learning Multi-Task Loco-Manipulation Whole-Body Control for Quadruped Robot With Arm
Whole-body loco-manipulation for quadruped robots with arms remains a challenging problem, particularly in achieving multi-task control. To address this, we propose MLM, a reinforcement learning framework driven by both real-world and simulation data. It enables a six-DoF robotic arm–equipped quadruped robot to perform whole-body loco-manipulation for multiple tasks autonomously or under human teleoperation. To address the problem of balancing multiple tasks during the learning of loco-manipulation, we introduce a trajectory library with an adaptive, curriculum-based sampling mechanism. This approach allows the policy to efficiently leverage real-world collected trajectories for learning multi-task loco-manipulation. To address deployment scenarios with only historical observations and to enhance the performance of policy execution across tasks with different spatial ranges, we propose a Trajectory-Velocity Prediction policy network. It predicts unobservable future trajectories and velocities. By leveraging extensive simulation data and curriculum-based rewards, our controller achieves whole-body behaviors in simulation and zero-shot transfer to real-world deployment. Ablation studies in simulation verify the necessity and effectiveness of our approach, while real-world experiments on a Go2 robot with an Airbot robotic arm demonstrate the policy's good performance in multi-task execution.
- 3 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-12
Development of a Neural Network-Based Compensatory Enhanced-Hammerstein Modeling Strategy for Piezoelectric System With Hysteresis
The accurate modeling of the piezoelectric actuator is fundamental to guarantee the rapidity and accuracy of positioning control for micro-nano manipulation. Nevertheless, the high-accuracy modeling of input voltage and driving displacement is difficult because the piezoelectric actuator has the creep, vibration dynamics, nonlinear hysteresis, and system uncertainty. In this paper, a Compensatory Nonlinear-Linear-Linear Hammerstein (CNLLH) modeling strategy considering creep, vibration dynamics, hysteretic nonlinearity, and uncertainty compensation for piezoelectric system is proposed to achieve an accurate input-output model for control design, where the input is voltage signal, the output is driving displacement. This model structure is a nonlinear-linear-linear-nonlinear block, where the first nonlinear model is used to describe hysteretic nonlinear behavior, the two linear models are used to describe creep and vibration characteristics, the second nonlinear model with a neural network is used to compensate the system uncertainty. The corresponding CNLLH model identification algorithm is developed based on the singular value decomposition, regularized least square method, and gradient descent method. The proposed CNLLH strategy can realize the simultaneous identification of four different characteristics only based on a set of experimental data without the decoupled experiment design. Experiments are conducted on the piezoelectric system to show the correctness of the proposed CNLLH strategy.
- 3 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-12
Development of a Neural Network-Based Compensatory Enhanced-Hammerstein Modeling Strategy for Piezoelectric System With Hysteresis
The accurate modeling of the piezoelectric actuator is fundamental to guarantee the rapidity and accuracy of positioning control for micro-nano manipulation. Nevertheless, the high-accuracy modeling of input voltage and driving displacement is difficult because the piezoelectric actuator has the creep, vibration dynamics, nonlinear hysteresis, and system uncertainty. In this paper, a Compensatory Nonlinear-Linear-Linear Hammerstein (CNLLH) modeling strategy considering creep, vibration dynamics, hysteretic nonlinearity, and uncertainty compensation for piezoelectric system is proposed to achieve an accurate input-output model for control design, where the input is voltage signal, the output is driving displacement. This model structure is a nonlinear-linear-linear-nonlinear block, where the first nonlinear model is used to describe hysteretic nonlinear behavior, the two linear models are used to describe creep and vibration characteristics, the second nonlinear model with a neural network is used to compensate the system uncertainty. The corresponding CNLLH model identification algorithm is developed based on the singular value decomposition, regularized least square method, and gradient descent method. The proposed CNLLH strategy can realize the simultaneous identification of four different characteristics only based on a set of experimental data without the decoupled experiment design. Experiments are conducted on the piezoelectric system to show the correctness of the proposed CNLLH strategy.
- 29 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Electronics
- Vol. 72 (2) , 1743-1752
Skill Latent Space Based Multigait Learning for a Legged Robot
Using proprioception to reproduce multiple gaits is nontrivial for legged robots, especially when encountering different terrains and velocity commands. Recently, reinforcement learning (RL) has been utilized to design powerful blind locomotion controllers. However, many RL-based blind locomotion methods are studied based on single gait. In this work, we propose an end-to-end training framework capable of learning multiple gaits for a quadruped robot. A latent space is constructed concurrently by a gait encoder and a gait generator, which helps the robot to reuse multiple gait skills to achieve adaptive gait (AG) behaviors. The trained controller enables smooth transitions between gaits and generates an AG. This means that the robot can effectively apply different gaits based on the current state and velocity command. To learn natural behaviors for multiple gaits, the reward explicitly constructed from gait parameters and the reward implicitly constructed from conditional adversarial motion priors together form the gait-dependent reward, which is subsequently added to the total reward. In the experiment, we demonstrate good performance of the multiple gaits control on a quadruped robot with only proprioceptive sensors.
- 29 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Electronics
- Vol. 72 (2) , 1743-1752
Skill Latent Space Based Multigait Learning for a Legged Robot
Using proprioception to reproduce multiple gaits is nontrivial for legged robots, especially when encountering different terrains and velocity commands. Recently, reinforcement learning (RL) has been utilized to design powerful blind locomotion controllers. However, many RL-based blind locomotion methods are studied based on single gait. In this work, we propose an end-to-end training framework capable of learning multiple gaits for a quadruped robot. A latent space is constructed concurrently by a gait encoder and a gait generator, which helps the robot to reuse multiple gait skills to achieve adaptive gait (AG) behaviors. The trained controller enables smooth transitions between gaits and generates an AG. This means that the robot can effectively apply different gaits based on the current state and velocity command. To learn natural behaviors for multiple gaits, the reward explicitly constructed from gait parameters and the reward implicitly constructed from conditional adversarial motion priors together form the gait-dependent reward, which is subsequently added to the total reward. In the experiment, we demonstrate good performance of the multiple gaits control on a quadruped robot with only proprioceptive sensors.
- 10 January 2023
- Published by Wiley in Journal of Mechanical Design
- Vol. 145 (5) , 1-14
A Modeling Method for a 6-SPS Perpendicular Parallel Micro-Manipulation Robot Considering the Motion in Multiple Nonfunctional Directions and Nonlinear Hysteresis
Precision motion systems that call for an ultrahigh resolution are widely used in ultrahigh-precision engineering applications. In this article, a 6-SPS perpendicular compliant parallel micro-manipulation robot is presented. It is difficult to establish the mapping relationship between the applied voltage and end-effector pose due to the existence of compliant joints and piezoelectric actuators. A modeling method for a 6-SPS perpendicular compliant parallel micro-manipulation robot considering the motion in nonfunctional directions and hysteresis effect is proposed to reflect the characteristic from input voltage to output pose. The motion of the spherical compliant joint is regarded as a spatial six degrees-of-freedom mechanism. This method considers the motion errors in multiple nonfunctional directions of the compliant joints. The internal force and torque are also taken into account based on the spatial force and moment balance condition. The slow, fast dynamic modes, and hysteresis nonlinearity behavior are described using a series and parallel hybrid model structures. The kinematic model and compliance model without considering the motion in nonfunctional directions are conducted for comparison. Simulation results show that the presented modeling method has higher modeling accuracy. Finally, experiments on a 6DOF compliant parallel micro-manipulation robot reveal the fine modeling performance of the developed method for the precision motion systems in the presence of intrinsic hysteresis effect and motion error of compliant joints.
- 10 January 2023
- Published by Wiley in Journal of Mechanical Design
- Vol. 145 (5) , 1-14
A Modeling Method for a 6-SPS Perpendicular Parallel Micro-Manipulation Robot Considering the Motion in Multiple Nonfunctional Directions and Nonlinear Hysteresis
Precision motion systems that call for an ultrahigh resolution are widely used in ultrahigh-precision engineering applications. In this article, a 6-SPS perpendicular compliant parallel micro-manipulation robot is presented. It is difficult to establish the mapping relationship between the applied voltage and end-effector pose due to the existence of compliant joints and piezoelectric actuators. A modeling method for a 6-SPS perpendicular compliant parallel micro-manipulation robot considering the motion in nonfunctional directions and hysteresis effect is proposed to reflect the characteristic from input voltage to output pose. The motion of the spherical compliant joint is regarded as a spatial six degrees-of-freedom mechanism. This method considers the motion errors in multiple nonfunctional directions of the compliant joints. The internal force and torque are also taken into account based on the spatial force and moment balance condition. The slow, fast dynamic modes, and hysteresis nonlinearity behavior are described using a series and parallel hybrid model structures. The kinematic model and compliance model without considering the motion in nonfunctional directions are conducted for comparison. Simulation results show that the presented modeling method has higher modeling accuracy. Finally, experiments on a 6DOF compliant parallel micro-manipulation robot reveal the fine modeling performance of the developed method for the precision motion systems in the presence of intrinsic hysteresis effect and motion error of compliant joints.