CV
Changda Tian
- Birth of Date: 11/19/1997
- Email: tianchangda97@gmail.com
- Cell: GR +30 6973678772 CN +86 18916906856
- Address： Nik. Plastira 100, Iraklio 700 13, Greece.
AREAS OF INTERESTS
- Robotics, Reinforcement Learning
EDUCATION
- Foundation for Research and Technology – Hellas (FORTH) Greece & Computer Science Department, University of Crete 09/2023 to now
- Major: Robotics + Computer Science
- School of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University, Shanghai, China 09/2016 to 06/2020
- Major: Automation + Computer Science
- Zhiyuan College, Shanghai Jiao Tong University 09/2016 to 06/2020
- Zhiyuan Honors Program of Engineering (Top 5%)
- School of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University, Shanghai, China 09/2020 to present
- Major: Control Engineering
SKILLS
- Programming: C/C++, Python
- Robotic systems: ROS
- 3D modeling: Solidworks
- ML systems: Pytorch
- Physical Simulation Plantforms: Mujoco, CoppeliaSim
- Data Acquisition & Processing: Matlab, Mathematica
- FPGA Developing: Verilog HDL, Quartus II
- Modeling and analyzing control systems and design controllers
HONORS & AWARDS
- National Scholarship (1%) Oct.2017
- CASC Aerospace Science and Technology Award (5%) Oct.2018
- Zhiyuan Exellent Scholarship (10%) (4 times) Dec.2016/ Dec.2017 / Dec.2018 / Dec.2019
- B Prize of Shanghai Jiao Tong University (15%) (3 times) Oct.2017/ Oct.2018 / Oct.2019
- H Prize in 2019 MCM April.2019
- Second Prize in East China Area NXP Smart Car Competition August.2019
- Graduate Student Academic Scholarship (3 times) Nov.2020/ Nov.2021 / Nov.2022
- PhD of Marie Skłodowska-Curie Actions Sept.2023
LABORATORY EXPERIENCE
Manipulate Tools as Human: Learning Human Style by Adversarial Motion Priors Jan. 2023 - Sept. 2023
- Advisor: Yue Gao (Professor of AI Institute, Shanghai Jiao Tong University)
- Leverages a deep neural network to model the complex dynamics of tool and object manipulation.
- The network is trained using a combination of real-world data and synthetic data generated by an adversarial network.
- The adversarial network is designed to generate realistic motion trajectories that match the statistical properties of human motion, which are then used to augment the training data for the manipulation network.
Adversarial-based Algorithm for Motion Balancing Control of Legged Robots Dec. 2021-Now
- Advisor: Yue Gao (Professor of AI Institute, Shanghai Jiao Tong University)
- Separate legged robot balance control into two separate agents, stance and swing legs.
- Stance legs controlled with WBC method are responsible for generating acceleration to track target velocity. While swing legs controlled with RL method are responsible for balance the robot.
- Use adversarial training for stance and swing legs to control the legged robot track the desired velocity fast and stable.
Capability-based Locomotion Control of Legged Robots 2020.09-2022.05
- Advisor: Yue Gao (Professor of AI Institute, Shanghai Jiao Tong University)
- Propose a definition of the traverse capability of legged robots, that is the traverse success rate of the robot, which is relevant to its surrounding terrain, topology and motion controller.
- Generate plenty of locomotion tasks in simulation to train our legged robot, Qingzhui. A layered control strategy maximizes its capability, at the same time learns its capability.
- Use the learned capability model to optimize long-range locomotion guidance algorithms
Designing Skating and Skiing Robot for Beijing Winter Olympic Torch Relay 2020.11-2022.02
- Advisor: Feng Gao (Professor of Mechanical Engineering School, Shanghai Jiao Tong University)
- Modify hexapod robots for skating and skiing. Design control algorithms for skating and skiing robots.
- Design environment perception and remote-control framework for skating and skiing robots.
- Conducted outdoor skating and skiing experiments in real skating and skiing resorts.
Designing Locomotion Controller for Six-Legged Robot 2019.09-2020.04
- Advisor: Yue Gao (Professor of Automation Department, Shanghai Jiao Tong University)
- Design a top-level controller based on reinforcement learning for path planning and a low-level controller based on trajectory optimization for gait and feet-pos planning of a hexapod.
- Use sim-to-real method, we implement the actions of the robot in simulation. Doing our training part in our simulation environment. And we use the trained model to run the actual hexapod robot
A Framework Based on Rule Reasoning and Syntactic Schema Embedding 2018.10-2019.01
- Advisor: Xinbing Wang (Professor of Computer Science Department, Shanghai Jiao Tong University)
- Implemented a core logic connection part of a Knowledge Based Question Answer (KBQA) system. First, use CRFBiLSTM to extract the key information of a query, including the subject and predicate and also label the key information. Then, use the labels of the key information to search in the knowledge graph, try to find a linked graph including all the information that extracted. Then, by the knowledge graph, we can derive the answer of the query.
- Had a patent about the KBQA system, Patent No: CN201910314357.X.
Fundamental Implementation of Neural Network on FPGA 2018.07-2018.09
- Advisor: Azita Emami (Professor of Moore Lab, California Institute of Technology)
- Found a way to represent floating point numbers in FPGA: Dividing the binary number to 3 parts, indicating the sign, integer and the decimal part. Each part can be adjusted by the actual accuracy requirements and the length limit.
- Implemented some basic matrix multipliers in FPGA using Verilog HDL. Then, with these multipliers, I made some activation functions that can be used in neural networks in FPGA.
- Synthesized the multipliers and the activation functions together and implemented a basic Deep Neural Network in Quartus II platform and programmed a basic neural network in Intel Stratix FPGA.
PUBLICATIONS
Chenyu Yang, Changda Tian, Qingshan Yao and Yue Gao. Sim-to-Real: Designing Locomotion Controller for Six-Legged Robot. In IEEE International Conference on CYBER Technology in Automation, Control, and Intelligent Systems (IEEE-CYBER), July 2019.
Changda Tian and Yue Gao. CapPlanner: Adaptable to Various Topology and Locomotion Capability for Hexapod Robots. In IEEE International Conference on Robotics and Biomimetics (IEEE-ROBIO), December 2022.
Changda Tian and Yue Gao. Learning Capability to Enhance Locomotion Control and Planning for Legged Robots. Submitted to IEEE RAM and under review, Sept 2023.
Ziqi Ma, Changda Tian and Yue Gao. Manipulate as Human: Learning Task-oriented Manipulation Skills by Adversarial Motion Priors. Submitted to IEEE ICRA 2024 and Under Review, Sept 2023.
