A5034142902- Robotic Locomotion and Control
- Robot Manipulation and Learning
- Reinforcement Learning in Robotics
- Prosthetics and Rehabilitation Robotics
- Robotic Path Planning Algorithms
- Modular Robots and Swarm Intelligence
- Muscle activation and electromyography studies
- Biomimetic flight and propulsion mechanisms
- Soft Robotics and Applications
- Robotics and Sensor-Based Localization
- Robotics and Automated Systems
- Greenhouse Technology and Climate Control
- Spine and Intervertebral Disc Pathology
- Anesthesia and Pain Management
- Viral Infectious Diseases and Gene Expression in Insects
- Stroke Rehabilitation and Recovery
- Virology and Viral Diseases
- Social Robot Interaction and HRI
- Anomaly Detection Techniques and Applications
- Hand Gesture Recognition Systems
- Advanced Image and Video Retrieval Techniques
- Teleoperation and Haptic Systems
- Human Pose and Action Recognition
- Acupuncture Treatment Research Studies
- Robotic Mechanisms and Dynamics
University of San Diego
2025
Massachusetts Institute of Technology
2023
University of California, Berkeley
2022
Guangzhou University of Chinese Medicine
2022
Nankai University
2019-2021
Shenyang University
2021
We address the problem of enabling quadrupedal robots to perform precise shooting skills in real world using reinforcement learning. Developing algorithms enable a legged robot shoot soccer ball given target is challenging that combines motion control and planning into one task. To solve this problem, we need consider dynamics limitation stability during dynamic robot. Moreover, hard-to-model deformable rolling on ground with uncertain friction desired location. In paper, propose...
Wearable robots like ankle exoskeletons have demonstrated the capability to enhance human mobility and reduce biological efforts of locomotion. The type assistance provided by could influence lower leg muscle activities during walking. This article aimed systematically evaluate under different exoskeleton conditions. We measured multiple electromyography-based metrics five muscles, while participants walked with an on a treadmill. Nine conditions, which combined three peak times (46%, 49%,...
DribbleBot (Dexterous Ball Manipulation with a Legged Robot) is legged robotic system that can dribble soccer ball under the same real-world conditions as humans. We identify key challenges of in-the-wild manipulation, including variable motion dynamics and perception using body-mounted cameras. To overcome these challenges, we propose domain task specification for learning viable dribbling behaviors in simulation transfer to real fields. Our provides promising evidence current robots are...
Abstract The greenhouse is a multivariate, nonlinear, time-varying and time-lag system. establishment of an accurate temperature prediction model can effectively ensure the effectiveness intelligent control. In order to solve problem difficulty in modeling mechanism greenhouses, this paper proposes method based on improved Support Vector Regression establish prediction, PSO algorithm used optimize hyper-parameters SVR. Using real data verify proposed model. Experimental results show that...
Abstract Background Lumbar disc herniation is seen in 5–15% of patients with lumbar back pain and the most common spine disorder demanding surgical correction. Spinal surgery one effective management for these patients. However, current techniques still present complications such as chronic 10–40% all who underwent surgery, which has a significant impact on patients’ quality life. Research studies have shown that transcutaneous electrical acupoint stimulation (TEAS) may reduce cumulative...
Can we enable humanoid robots to generate rich, diverse, and expressive motions in the real world? We propose learn a whole-body control policy on human-sized robot mimic human as realistic possible. To train such policy, leverage large-scale motion capture data from graphics community Reinforcement Learning framework. However, directly performing imitation learning with dataset would not work robot, given large gap degrees of freedom physical capabilities. Our method Expressive Whole-Body...
We study the problem of mobile manipulation using legged robots equipped with an arm, namely loco-manipulation. The robot legs, while usually utilized for mobility, offer opportunity to amplify capabilities by conducting whole-body control. That is, can control legs and arm at same time extend its workspace. propose a framework that conduct autonomously visual observations. Our approach, \ourFull~(\our), is composed low-level policy all degrees freedom track end-effector manipulator position...
Combining the mobility of legged robots with manipulation skills arms has potential to significantly expand operational range and enhance capabilities robotic systems in performing various mobile tasks. Existing approaches are confined imprecise six degrees freedom (DoF) possess a limited arm workspace. In this paper, we propose novel framework, RoboDuet, which employs two collaborative policies realize locomotion simultaneously, achieving whole-body control through interactions between each...
Motivated towards performing missions in unstructured environments using a group of robots, this paper presents reinforcement learning-based strategy for multiple quadrupedal robots executing collaborative manipulation tasks. By taking target position, velocity tracking, and height adjustment into account, we demonstrate that the proposed enables four manipulating payload to walk at desired linear angular velocities, as well over challenging terrain. The learned policy is robust variations...
Controlling contact forces during interactions is critical for locomotion and manipulation tasks. While sim-to-real reinforcement learning (RL) has succeeded in many contact-rich problems, current RL methods achieve forceful implicitly without explicitly regulating forces. We propose a method training policies direct force control requiring access to sensing. showcase our on whole-body platform of quadruped robot with an arm. Such enables us perform gravity compensation impedance control,...
This paper proposes to solve the problem of Vision-and-Language Navigation with legged robots, which not only provides a flexible way for humans command but also allows robot navigate through more challenging and cluttered scenes. However, it is non-trivial translate human language instructions all low-level leg joint actions. We propose NaVILA, 2-level framework that unifies Vision-Language-Action model (VLA) locomotion skills. Instead directly predicting actions from VLA, NaVILA first...
DribbleBot (Dexterous Ball Manipulation with a Legged Robot) is legged robotic system that can dribble soccer ball under the same real-world conditions as humans (i.e., in-the-wild). We adopt paradigm of training policies in simulation using reinforcement learning and transferring them into real world. overcome critical challenges accounting for variable motion dynamics on different terrains perceiving body-mounted cameras constraints onboard computing. Our results provide evidence current...
Knowledge of terrain's physical properties inferred from color images can aid in making efficient robotic locomotion plans. However, unlike image classification, it is unintuitive for humans to label patches with properties. Without labeled data, building a vision system that takes as input the observed terrain and predicts remains challenging. We present method overcomes this challenge by self-supervised labeling captured robots during real-world traversal property estimators trained...
Robot navigation under dynamic environment is considered as one of the key indications machine's intelligence. This paper proposed a program learning algorithm for robot navigation, so to efficiently increase intelligence machine. The Tolman mouse maze experiment was expanded help develop intuitive intelligence, that it can lean regularity change. Firstly, efficient map constructed by SOM based on cloud point generated randomly according environment. Then will automatically choose how reach...
Convolutional neural network based on
We address the problem of enabling quadrupedal robots to perform precise shooting skills in real world using reinforcement learning. Developing algorithms enable a legged robot shoot soccer ball given target is challenging that combines motion control and planning into one task. To solve this problem, we need consider dynamics limitation stability during dynamic robot. Moreover, hard-to-model deformable rolling on ground with uncertain friction desired location. In paper, propose...