Xiaopeng Ji

Xiaopeng Ji

Auto-generated profile, based on publicly available information

Xiaopeng Ji

State Key Lab of CAD&CG, Zhejiang University, Hangzhou, China

Research Focus & Metrics

Academic Career Length

14Years (2013-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
8 November 2022 in IEEE Robotics and Automation Letters
Verified
Journal Article
Closed access
PZPeng Zhai
Peng Zhai
THTaixian Hou
Taixian Hou
XJXiaopeng Ji
Xiaopeng Ji
ZDZhiyan Dong
Zhiyan Dong
LZLihua Zhang
Lihua Zhang
Reinforcement learning needs to learn policies through trial and error. The unstable policies in the early stage of training make it expensive (and time-consuming) to train directly in the real environment, which may cause disastrous consequences. The popular solution is to use the simulator to train the policy and deploy it in a real environment. However, the modeling error and external disturbance between the simulation and the real environment may fail the physical deployment, resulting in the sim2real transfer problem. In this letter, we propose a novel robust adversarial reinforcement learning framework, which uses the ensemble training of multi-adversarial agents that can adaptively adjust adversaries' strength to enhance RL policy's robustness. More specifically, we take the accumulative reward as feedback and construct a PID controller to adjust the adversary's output magnitude to perform the adversarial training well. Experiments in the simulated and the real environment show that our algorithm improves the generalization ability of the policy for the modeling error and the uncertain disturbance simultaneously, outperforming the next best prior methods across all domains. The algorithm was further proven to be effective in a sim2real transfer task through the load experiment of a real racing drone, and the tracking performance is better than the PID-based flight controller.

6 December 2013 in PLOS ONE
Verified
Journal Article
Open access
PDF file
HJHui Jiao
Hui Jiao
ZXZhuodong Xu
Zhuodong Xu
LWLebin Wu
Lebin Wu
ZCZhaoping Cheng
Zhaoping Cheng
XJXiaopeng Ji
Xiaopeng Ji
CMChen Meng
Chen Meng
To assess the feasibility of low-dose prospective ECG-gated dual-source CT (DSCT) in detecting airway anomalies in pediatric patients with cardiovascular anomalies compared with flexible tracheobronchoscopy (FTB). 33 pediatrics with respiratory symptoms who had been revealed cardiovascular anomalies by transthoracic echocardiography underwent FTB and contrast material–enhanced prospective ECG-triggering CT were enrolled. The study was approved by our institution review board and written informed consent was obtained from all patients’ guardian. DSCT examinations were performed to detect cardiovascular abnormalities using weight-adjusted low–dose protocol. Two radiologists independently performed CT image analysis. The FTB reports were reviewed by an experienced pulmonologist. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy of DSCT in the detection of airway anomalies were assessed. The tracheobronchial stenoses revealed on FTB were graded. Effective radiation dose was calculated. Thirty cases were diagnosed with tracheobronchial narrowing and/or abnormality in 33 patients by FTB, while 3 patients had normal FTB findings. Twenty-eight cases were diagnosed with airway anomalies by CT, of which 27 were correct positive. 3 patients with normal findings at CT had findings of tracheobronchial narrowing due to tracheobronchomalacia at inspiration at FTB. Sensitivity and specificity of CT were 90.0% (95% CI: 72.3%, 97.4%) and 66.7% (95% CI: 12.5 %, 98.2 %), respectively. PPV and NPV were 96.4% (95% CI: 79.8 %, 99.8%) and 40.0% (95% CI: 7.3%, 83.0%), respectively. Overall accuracy of DSCT in detecting airway anomalies in pediatrics with cardiovascular anomalies was 87.9% (95% CI: 74.5%, 97.6%). In grading of tracheobronchial stenosis, images from CT correlated closely (r = 0.89) with those of FTB. Mean effective dose was 0.60±0.20 mSv. In pediatric patients, ECG-triggered CT to evaluate congenital cardiovascular anomalies can also be used to diagnose and characterize fixed airway involvement in relation to the vascular structures.

1 September 2015
Verified
Conference Paper
Closed access
XJXiaopeng Ji
Xiaopeng Ji
JCJun Cheng
Jun Cheng
DTDapeng Tao
Dapeng Tao
With the promptly growing population of the low-cost Microsoft Kinect sensor, action recognition, which is a hard yet important problem in computer vision, has been received substantial attention. However, most existing approaches in action recognition spend much time on feature detection even though these methods can achieve high recognition rates. In this paper, we propose a local mean spatio-temporal feature (LMSF) to speed up depth image based action recognition. In particular, we solve the problem from three aspects: (1) associate the 4D normals by a local mean spatio-temporal neighborhood; (2) extract motion frames by detecting the differences between consecutive frames; (3) reduce redundant normals extracted from depth cloud points by sparse coding. The proposed approach is tested on two public benchmark datasets, i.e., MSRAction3D and MSRGesture3D. Experimental results demonstrate the advantages of our improvement method and the state-of-the-art performance on processing speed.

3 October 2024 in IEEE Robotics and Automation Letters
Verified
Journal Article
Closed access
PZPeng Zhai
Peng Zhai
XWXiaoyi Wei
Xiaoyi Wei
THTaixian Hou
Taixian Hou
XJXiaopeng Ji
Xiaopeng Ji
ZDZhiyan Dong
Zhiyan Dong
LZLihua Zhang
Lihua Zhang
Reinforcement learning (RL) can generate high-performance control policies for complex tasks in simulation through an end-to-end approach. However, the RL policy is not robust to uncertainties caused by modeling mismatch between simulation and real environments, making it difficult to transfer to the real world. In response to the above challenge, this letter introduces a lightweight and efficient robust RL algorithm. The algorithm transforms the optimization objective of the adversary from a long-term cumulative reward to a short-term reward, making the adversary focus on the performance in the near future. Additionally, the adversarial actions are projected onto a finite subset within the perturbation space using projected gradient descent, effectively constraining the adversary's strength and obtaining more robust policies. Extensive experiments in both simulated and real environments show that our algorithm improves the generalization ability of the policy for the modeling mismatch, outperforming the next best prior methods across almost all environments.