I received my Master’s and undergraduate degrees from the Beijing Institute of Technology(BIT), and was a member of the Robot Vision Group at the the Beijing Advanced Innovation Center for Intelligent Robots and Systems(BAICIRS). My background includes working on multiple vision-based robot projects, including Robotic Bionic Eyes, UR Robot Arm, and AGV. My research areas of interest include Robotics, Active SLAM, 3D Vision, and Reinforcement Learning, with a current focus on perception-based control, robot learning, and cognitive robotics. My ambition is to develop a practical cognitive robot prototype.
Head stabilization for Robotic Bionic Eyes
I designed a Disturbance Decomposition and Learning (DDL) control system for a robotic bionic eyes platform. It separates disturbance in the nonlinear system and compensates for it without complicated dynamic modeling, and leverages a learning-based approach to address nonlinear errors. The results showed improved control accuracy compared to state-of-the-art methods, both in simulation and real-world experiments. The work has been submitted to TMECH.
Autonomous Exploration with Active Camera Control
Aug. 2021-Jun. 2022
I designed an algorithm for active exploration that utilizes multimodal sensory fusion and active camera control. Simulations demonstrate that compared to traditional algorithms, this approach increases map coverage by 10-20% and improves exploration efficiency by 33%. The work has been accepted for publication in the CCDC 2022. And the work has also been extended to unmanned aerial vehicle (UAV) platforms.
Gap-based Topological Mapless Maze Navigation
Nov. 2021-Jun. 2022
I collaborated with lab member Zakir on a project focused on mapless topological navigation. The paper introduced a new approach, named “Follow the Nearest-to-final goal and Unvisited Gap (FNUG),” for navigating imperfect unknown mazes. The algorithm is computationally efficient and can handle loops and dead ends with its revisit check feature. The paper was published in the RA-L 2022.
Reinforcement Learning based UAV control
Sep. 2019-Aug. 2020
I co-authored a paper on using reinforcement learning for unmanned aerial vehicle control. The PPO algorithm of reinforcement learning was utilized to tune PID controller gains. The paper was published in the CCC 2020.
UR Robot Arm based Mobile Platform for Manipulation Task (with Huawei)
Sep. 2020-Jun. 2022
I was part of the team that designed the full system framework for a robot system based on a UR robot arm and ROS. I played a significant role in developing modules such as manipulator planning and control, object detection, pose estimation, visual servoing, and socket communication. I was responsible for conducting real robot tests. Additionally, I created a corresponding simulation platform using Gazebo.
Sep. 2019-Aug. 2020, Active Mapping
The goal of this project was to create a dynamic SLAM framework using bionic eyes for robots to actively explore and map indoor environments. The optimal exploration path was determined using Information Gain computed from the real-time map, and a MPC controller was utilized for precise tracking of the local trajectory, leading to successful complete mapping of a laboratory and corridor.
Sep. 2019-Aug. 2020, Camera-IMU Time-Synchronization for Robotic Bionic Eyes based SLAM
The objective of this project was to enhance the accuracy of visual-inertial SLAM on a bionic eye platform by creating a camera-IMU time synchronization module. The embedded development board was equipped with hardware trigger synchronization and IMU and camera time error measurement. The resulting synchronized images and IMU data were gathered and shared via ROS, leading to a more robust visual-inertial SLAM system. A patent has been granted for this work.
Sep. 2020-Aug. 2021, Stereo Depth Estimation for Robotic Bionic Eyes
The objective of this project was to create a depth estimation module for the bionic eyes’ cameras, enabling 3D reconstruction capabilities. The intrinsic and extrinsic parameters were calibrated, and an image acquisition program for CCD cameras was developed. Both traditional methods and machine learning-based approaches were utilized.
Sep. 2019-Aug. 2020, RGB-D based Indoor Dense 3D Reconstruction and Navigation
The aim of this project was to achieve high-precision indoor 3D reconstruction and navigation using an RGB-D camera. A TSDF-based dense mapping algorithm was employed for detailed map construction, and LIDAR was utilized for localizing and navigating the mobile platform.
Dec. 2020-Jan. 2021, Infrared Camera based Human Body Temperature Measurement
The goal of this project was to develop temperature measurement software for a small spherical robot using an infrared camera.
Teaching Assistant Experience
Nov. 2020-Jan. 2021, Python & Deep Learning Basics (Undergraduate course)
This course is set by School of Mechatronical Engineering at BIT.
Publications & Patents
- G. Siyuan, C. Xiaopeng, Z. Weizhong, et al. “DDL: Disturbance Decomposition and Learning Based Head Stabilization for Robotic Bionic Eyes, “ (Submitted to IEEE/ASME Transactions on Mechatronics).
- G. Siyuan, C. Xiaopeng, Z. Weizhong, et al. “Multimodal Perception based Autonomous Exploration with Active Camera Control in Unknown Environments,” (Accepted by CCDC 2022)
- U. Zakir, C. Xiaopeng, G. Siyuan, Y. Xu and S. Muhammad, “FNUG: Imperfect mazes traversal based on detecting and following the nearest-to-ﬁnal-goal and unvisited gaps,” in IEEE Robotics and Automation Letters, doi: 10.1109/LRA.2022.3151393.
- C. Xiaopeng, L. Yanyang, X. Yang, G. Siyuan, M. Siyan, U. Zakir, “Kinematic Calibration of a Laser Tracker based on Nonlinear Optimization of a Refined Geometric Error Model,” Measurement (2022): 110672.
- Z. Qingqing, T. Renjie, G. Siyuan, C. Xiaopeng, Z. Weizhong, “A PID Gain Adjustment Scheme Based on Reinforcement Learning Algorithm for a Quadrotor,” 2020 39th Chinese Control Conference (CCC), 2020, pp. 6756-6761.
- Q. Huang, C. Xiaopeng, G. Siyuan, et al. Method and Apparatus for time-synchronization of bionic eyes’ multi-channel IMUs and cameras, ZL201910777841.6 (Chinese Patent).
- C. Xiaopeng, Q. Huang, G. Siyuan, et al. Method and Apparatus for Pose Planar Constraining on the Basis of Planar Feature Extraction, US-2021-0192271-A1 (U.S. Patent).