A5031185739- Reinforcement Learning in Robotics
- Robotic Locomotion and Control
- Osteoarthritis Treatment and Mechanisms
- Prosthetics and Rehabilitation Robotics
- Rheumatoid Arthritis Research and Therapies
- Natural product bioactivities and synthesis
- Traditional Chinese Medicine Analysis
- Traffic control and management
- Robot Manipulation and Learning
- Musculoskeletal pain and rehabilitation
- Smart Grid Energy Management
- Fibromyalgia and Chronic Fatigue Syndrome Research
- Adaptive Dynamic Programming Control
- Pharmacological Effects of Natural Compounds
- Orthopedic Infections and Treatments
- Bone health and osteoporosis research
- Bone Metabolism and Diseases
- Bone and Joint Diseases
- Data Stream Mining Techniques
- Soft tissue tumor case studies
- Inflammatory mediators and NSAID effects
- Transportation and Mobility Innovations
- Sarcoma Diagnosis and Treatment
- Software Testing and Debugging Techniques
- Robotic Path Planning Algorithms
Beijing University of Chinese Medicine
2022-2025
Nanjing University
2017-2024
This paper addresses the problem of legged locomotion in unstructured environments, and a novel Hierarchical multi-contact motion planning method for hexapod robots is proposed by combining Free Gait Deep Reinforcement Learning (HFG-DRL). We structurally decompose complex free gait task into path discrete state space continuous space. Firstly, Soft Q-Network (SDQN) used to obtain global prior information Path Planner (PP). Secondly, (FGP) sequence. Finally, based on PP FGP, Center-of-Mass...
Current clinical studies on femoral head necrotic lesions primarily focus the medial and lateral regions, while detailed MRI-based methods to evaluate relationship between anterior or posterior necrosis collapse remain lacking. By defining positions of in MRI, a method was proposed for rapid prognosis assessment based location. A retrospective analysis conducted TSE sequence T1W1 coronal plane images from 200 cases necrosis. The frequency appearing each MRI layer statistically analyzed...
Abstract Background Given the high cost of endoscopy in gastric cancer (GC) screening, there is an urgent need to explore cost-effective methods for large-scale prediction precancerous lesions (PLGC). We aim construct a hierarchical artificial intelligence-based multimodal non-invasive method pre-endoscopic risk provide tailored recommendations endoscopy. Methods From December 2022 2023, screening study was conducted Fujian, China. Based on traditional Chinese medicine theory, we...
Background Accurate preoperative histological stratification (HS) of intracranial solitary fibrous tumors (ISFTs) can help predict patient outcomes and develop personalized treatment plans. However, the role a comprehensive model based on clinical, radiomics deep learning (CRDL) features in HS ISFT remains unclear. Purpose To investigate feasibility CRDL magnetic resonance imaging (MRI) ISFT. Study Type Retrospective. Population Three hundred ninety‐eight patients from Beijing Tiantan...
Safe and autonomous obstacle avoidance plays an important role in the navigation control of hexapod robots. In this paper, we combine method reinforcement learning with fuzzy to achieve for a robot complex environments. A Q-learning algorithm is first presented approach proposed using Fuzzy regarding specific requirements robot. Then, implemented real system that uses ultrasonic sensors detect obstacles unknown environment learns optimal policy avoid obstacles. Several groups experiments are...
Legged locomotion in a complex environment requires careful planning of the footholds legged robots. In this paper, novel Deep Reinforcement Learning (DRL) method is proposed to implement multi-contact motion for hexapod robots moving on uneven plum-blossom piles. First, formulated as Markov Decision Process (MDP) with specified reward function. Second, transition feasibility model robots, which describes state under condition satisfying kinematics and dynamics, turn determines rewards....
Rapid adaptation to the environment is long-term task of reinforcement learning. However, learning faces great challenges in dynamic environments, especially with continuous state-action spaces. In this paper, we propose a systematic Incremental Reinforcement Learning method via Performance Evaluation and Policy Perturbation (IRL-PEPP) improve adaptability algorithms environments spaces, which mainly includes three parts, i.e., performance evaluation, policy perturbation importance...
Legged locomotion in unstructured environments with static and dynamic obstacles is challenging. This paper proposes a novel Hierarchical Multi-Contact motion planning method Incremental Reinforcement Learning (HMC-IRL) that enables hexapod robots to pass through large-scale discrete complex local changes occurring. Firstly, hierarchical structure an information fusion mechanism are developed decompose multi-contact into two stages: the high level prior grid path low detailed COM foothold...
Background Osteoking has been extensively used for the treatment of knee osteoarthritis (KOA). However, it is lack high-quality evidence on clinical efficacy against KOA and comparison with that nonsteroidal anti-inflammatory drugs (NSAIDs). Aims To evaluate safety in treating KOA. Methods In current study, a total 501 subjects were recruited from 20 medical centers, divided into group ( n = 428) NSAIDs 73). The Propensity Score Matching method was to balance baseline data different groups....
A key skill for mobile robots is the ability to navigate efficiently through their environment, and reinforcement learning widely used in path planning robots. However, this algorithm has a slow convergence speed large number of iterations. There are few studies on how improve from perspective acquisition rule-based shallow-trial strategy. In biological world, animals depend own empirical knowledge when making planing. Humanity transcendental knowledge, which great help peoples navigation....
As a widely used machine learning method, reinforcement (RL) is very effective way to solve decision and control problems where skills are needed. In this paper, knowledge transfer method between multi-granularity models proposed for RL speed up the process adapt dynamic environments. The runs on naturally organized models, e.g., coarse-grained model fine-grained model. This constitutes architecture that bridges different granularity levels. (MGRL) approach related algorithms can scale well...