Bangguo Yu

Bangguo Yu

PhD Student

University of Groningen

Biography

When I stood on the field of ROBOCON competition in 2017’s, behind was the encouragement of teammates and in front of was the hope of victory. At that moment I fell in love with Robotics. As the major of automation student, I feel honor when I processed the part of robot, welded circuit board and debugged the program. Every coherent action of the robot is the most beautiful art in my eyes. And when I was familiar with the bottom-level knowledge of robotics, the high-level task of the robotic in perception and planning was my goal.

These movies “Edge of Tomorrow” and “Next” inspire me with the method of learning for the automatic robot. In my opinion, deep reinforcement learning can help robots achieve the same performance as movies by designing a reasonable reward. Parsing and utilizing effective environmental information of the 3D scene is not trivial due to the complexity of the 3D environment and the limited ability for reasoning about our visual world. Of course, the day has not yet come, and seems to be difficult, but it is because of its uncertainty that it’s worth studying.

Interests

  • Robotics
  • 3D Scene Graph
  • Target-driven Navigation

Education

  • PhD in Robotics, 2021.11 - Now

    University of Groningen

  • MEng in Control Engineering, 2018.09 - 2021.6

    Shandong University

  • BSc in Automation, 2014.09 - 2018.06

    Shandong University of Science and Technology

Skills

Robotics

ROS, PID Control, Sensor Filter, Navigation

SLAM

Mapping, Location, 3D scene graph

Learning

Reinforcement learning, Habitat platform

Programming

C++, Python, C, Matlab

Embedden System

STM32, K60, CAN, UART, IIC, SPI

Design

Altium Designer for hardware, SolidWorks for machine

Experience

 
 
 
 
 

Reserch Intern

DMAI

Aug 2019 – Jan 2020 Guangzhou, China
Responsibilities include:

  • Indoor 3D structured scene graph modeling
  • 3D object relationship detection
  • Semantic navigation demo based on 3D scene graph
 
 
 
 
 

Embedded engineer

Shandong University

May 2019 – Aug 2019 Shandong, China
Responsibilities include:

  • Designing embedded control board hardware and shell
  • Communication with server by CAN to control more than ten borad simultaneously
  • Control strong electricity (220V AC) using weak current (5V DC)
 
 
 
 
 

SLAM Algorithm engineer

Shandong University

Jul 2018 – Mar 2019 Shandong, China
Responsibilities include:

  • Reading the Cartographer designed by Google
  • The open loop optimization of SLAM based on apriltag to overcome the gallery with sparse features
  • Deploying the algorithm on the real robot

Award

The 14th China Graduate Electronic Design Contest

We use apriltag to achieve the object tracking and use ROS to display the tracking on the speacial IndustriPi.

National Undergraduate Electronic Design Contest

Our team designed a Ball and Plate control system with rgb camera. We used PID algorithm to set the velocity and position close loop, and achieved the control performance of ball’s movement.

The 5th and 6th Shandong Provincial Robot Competition

This is the preliminary for ROBOCON, and our team got the champion for two consecutive years in Shandong province.

The 16th and 17th China University Robot Competition (ROBOCON)

Our team contains more than twenty prople, divided in software, hardware circuit design and machine design. As the captain of our team, I am in charge of the recruiting, training, and all the research route.

Challenge Cup Technological Innovation Competition

An underwater robot was designed to detect the water quality and underwater searching.

The 5th Shandong Provincial University Robot Competition

Our team took part in the SYNCHRONOUS ROBOT and got the best grade in this item.

The 2th Shandong SCM application design Competition

Our team designed the smart car using 15 photoelectric switch sensors and got the best grade in this item.

National Undergraduate Smart Car Contest

We took play of the ‘double cars chasing’, the hardest item of this contest. This was the first contest in my college, and I learned ARM, coding, image processing and PID control system from the contest. We got the champion in Shandong province contest and the second prize in national contest.

Projects

*

Frontier Semantic Exploration for Visual Target Navigation

The priors represented by scene graph are incorporated in deep reinforcement learning model using R-GCN on Habitat platform.

Visual Target Navigation using 3D Scene Priors

The priors represented by scene graph are incorporated in deep reinforcement learning model using R-GCN on Habitat platform.

3D structured semantic scene graph

The semantic environment model of 3D structured scene graph for the implementation of robot task

Object tracking

Object tracking based on IndustriPi

The design of control board

Embedded board with control, hardware and shell

Cloud-based Open loop optimize SLAM

Open loop optimize SLAM based on Cartographer

Robocon robot

China University Robot Competition

Synchronous robot

How to run with a slide bar in two cars.

Ball and Plate Control System

National Undergraduate Electronocs Design Contest

Double Smart Cars

Double cars chase and overtake on the road.

Recent Publications

Visual Target Navigation using 3D Scene Priors

(Submitted in IROS2021) high-level human-robot interaction tasks, 3D semantic scene graphs

A Bottom-up Framework for Construction of Structured Semantic 3D Scene Graph

(Accepted by IROS2020) high-level human-robot interaction tasks, 3D semantic scene graphs