Past Projects

2012.1 ~ 2014.1, UMN

1. Search for People with Kinect using Multi-robot Reinforcement learning. (course project)

2. Simultaneous Localization and People Search with Kinect. (course project)

3. Line Features Based Simultaneous Localization and Mapping. (course project)

2009.1 ~ 2011.12, KYH

4. SLAMSAT (Searching and Tracking conditioned on SLAM)

2008.1~2008.12, ITRI

5. Intelligent Localization, Control and Navigation Technology

6. Intelligent underwater navigation and collision avoidance Technology

7. 3D vSLAM for auto parking (Project Leader)

2007.1~2007.12, ITRI

8. Dynamic Bayesian Network for Detection of Blind Spot and Parking of Vehicle (Project Leader)

9. Power Aided and Localization & Control Technology

10. Developing an electric control module of balanceable two wheel toy robot

2006.09~2006.12, MIT

11. Development of Simultaneous Localization and Mapping (SLAM) for Security Robot (International Cooperation Project with MIT CSAIL)

2006.1~2006.09, ITRI

12. mobile-platform control and application technologies

2005.1~2005.12, ITRI

13. Development of differential mobile platform

14. Development of omni-direction mobile platform

2002.6~2004.9, NTU

15. Design and Control of Quadruped Robot Using Integrated Analog and Digital Circuits

Introduction:

I had been involved in 15 robotic projects since 2002 in NTU, ITRI, MIT CSAIL, KYH and UMN .

I learned how to design mechanical parts of robots using CAD software (AutoCAD/ SoildWorks).

I learned how to control robots using Micro controller (ATMEL/Silicon Labs 8051), DSP (TI TMS320 28XX Series) and FPGA.

I learned how to program bayesian algorithms (Kalman Filter/Particle Filter/RBPF) for localization, mapping, people tracking and searching.

It turned out to be even more interesting than I thought it would be!

This video is to briefly describe past projects. Here is a larger version on youtube video (Recommended).

1. Search for People with Kinect using Multi-robot Reinforcement Learning [2012.1~2014.01, UMN]

Abstract:

Bayesian search is a technique to save thousands people every year in outdoor environment. However, searching for people in indoor environments is different. First, the robot has to build a map. Second, a robot has to localize itself. Third, people detection is not generally accurate. Fourth, to compute the best actions for the search is NP-hard problem due to the obstacles. Therefore, this research is about proposing an algorithm of search for people in an indoor environment using reinforcement learning. The experimental results demonstrate that this approach is able to reduce the uncertainty of search with reasonable computational performance.

2. Simultaneous Localization and People Search with Kinect [2012.1~2014.01, UMN]

Abstract:

Search is one of the oldest problems in AI Field. However, searching people in indoor environments is different from pervious search problems. First, a robot has to know its position. Second, the robot has to build map. Third, people detection is not accurate. Finally, there are a lot of obstacles. Therefore, this project is to implement a simultaneous localization and people search algorithm for a mobile robot with kinect.

3. Line Features Based Simultaneous Localization and Mapping [2012.1~2014.01, UMN]

Abstract:

In this project, a line feature based EKF SLAM is implemented. There are two assumptions for the environment. First of all, the environment is two dimensional and thus it could be described by line features. Secondly, there is no moving object in the environment. There are two assumptions for estimators. First, both of the probability density functions (pdf) of input signal and noise are Gaussian distributions. Second, the Bayesian networks of SLAM obey Markov chain. Because the most indoor environments can be modeled as line features easily, there are many approaches of line feature based SLAM [1] [2]. The approach in [2] is the main reference of this project.

4. SLAMSAT(Searching and Tracking conditioned on SLAM) [2009.5~2011.12, KYH]

Abstract:

SLAMSAT is an algorithm for a robot to localize itself, localize the environment, track the people in view and search the people out of sight. People searching and tracking (SAT) is a key technology for interactive robots since the tracked people are sheltered by environments frequently. For robots, it is a tracking problem given that the target is observable, but otherwise it is a searching problem. Traditional tracking algorithms may lead to divergent estimation of object position when moving objects are unobservable. Moreover, SAT conditioned on simultaneous localization and mapping (SLAM) is complex since it aims at estimating people position, robot position, and map under sensor uncertainty. Motivated by this, we propose a novel stream functions and Rao-Blackwellised particle filter based SAT algorithm. This laser based algorithm is conditioned on simultaneous localization and mapping (SLAMSAT) to search and track people. With this, the position of the targeted person sheltered by the environment can be successfully estimated by the virtual stream field in a mapping environment.

The red rectangle is estimated people position. The red circle is estimated people goal.

5. [Intelligent Localization, Control and Navigation Technology] [2008.1~2008.12, ITRI]

Abstract:

The trend of intelligent robotic systems and applications is moving from industrial robots toward service applications. Therefore, intelligent robot should be capable of dealing and interacting with environmental changes to meet users’ needs. This project proposes multi-sensor fusion technologies that aim to improve the disadvantages of single, expensive and unstable sensor measurements. And then establish localization, Control and Navigation technologies for robots in real and unstructured world.

Contributions:

I lead a team comprising 4 colleagues with Ph.D and master degrees to develop people tracking and SLAM for mobile robots, localization for underwater robots and auto-mapping for vehicles. Moreover, We applied 5 patents in probabilistic robotics domain [1] and [2]. For fusion of encoder and electrical comapss, I write a I2C interface to communicate with electrical compass using VHDL.

*This work is supported by Industrial Technology Research Institute (Mechanical and System Research Laboratories) in Taiwan.

6. [Intelligent Underwater Navigation and Collision Avoidance Technology] [2008.1~2008.12, ITRI]

Abstract:

For applications of intelligent underwater robotic systems, the robot should be capable of dealing and interacting with environmental changes to meet users’ needs. This project proposes multi-sensor fusion technologies that aim to improve the disadvantages of single, expensive and unstable sensor measurements. And then establish localization, Control and Avoidance collision technologies for underwater robots in real and unstructured world. The targets of the technologies are: (1) Robot with sonar and IMU can localize itself in the environment (2) Robot with sonar and IMU can simultaneously localize, finding obstacle, and avoid it.

Contributions:

I taught my team members how to estimate underwater localization using IMU and Sonar.

*This work is supported by Industrial Technology Research Institute (Mechanical and System Research Laboratories) in Taiwan.

7. [3D vSLAM for auto parking] [2008.1~2008.12, ITRI]

Abstract:

The project goal is to develop a mapping algorithm for parking of light weight electric vehicles.

Contributions:

I was the project leader and developed a gird-based mapping algorithm for vehicles.

*This work is supported by Industrial Technology Research Institute (Mechanical and System Research Laboratories) in Taiwan.

8. [Dynamic Bayesian Network for Detection of Blind spot and Parking of vehicle] [2007.1~2007.12, ITRI]

Abstract:

Object tracking of blind spot is a necessary technology of vehicle security for the development of intelligent vehicle. In order to track effectively the unknown moving object at blind spot, the project will adopt the newest AI technology , dynamic bayesian network(DBN), to estimate the position of vehicle, information of environment and dynamic state of the unknown moving object using a laser range finder and encoders. The system will remind drivers where the unknown moving object is, and then increase the traffic security.

9. [Power Aided and Localization and Control Technology] [2007.1~2007.12, ITRI]

Abstract:

The trend of intelligent robotic systems and applications is moving from industrial robots toward service applications. Therefore, intelligent robot should be capable of dealing and interacting with environmental changes to meet users’ needs. This project proposes multi-sensor fusion technologies that aim to improve the disadvantages of single, expensive and unstable sensor measurements. And then establish power aided and localization & Control technologies for robots in real and unstructured world. The targets of the technologies are: (1) The robot can stably climb up the slope and (2) The robot can localize itself accurately, and then complete the assigned tasks.

10. [Developing an Electric Control Module of Balanceable two wheel toy robot] [2007.1~2007.12, ITRI]

Abstract:

This project aims to develop a sensor fusion and a linear control technology to develop an electric control module of balanceable two wheel toy robot.

Contributions:

In order to estimate the angle of robot, I use Kalman Filter to fuse 3 axes-accelerometers and 1 axis-gyro scope. The CPU is DSP 28015. I write the firmware of DSP and VHDL of FPGA for catching encoder data. We applied a patent [1].

*This work is supported by Industrial Technology Research Institute (Mechanical and System Research Laboratories) in Taiwan.

Reference:

1. Kuo-Shih Tseng, Long-Der Chen, Ching-Yi Liu, “A Detection of External Force For Static and Moving Objects System and It’s Method”.

R.O.C. Patent No. 96150037 (pending)

11. [Mobile-platform control and Application Technologies] [2006.1~2006.09, ITRI]

Abstract:

In the past decades, software algorithm, hardware circuits and mechanical design were hard to integrate so robotic development was limited. For the purpose of fast accomplishing more competitive products by researchers all the world, main robotic research unit and Microsoft join in the robotic research and development. This project is to develop a General-Purpose Mobile Platform,called UBOT, researched and developed by ITRI/MSL.

Contributions:

I am a core member of UBOT team. I design the 1st version of motherboard,and write the VHDL of FPGA for communication Interface (I2C/ADC/IO) and parts of DSP codes.

*This work is supported by Industrial Technology Research Institute (Mechanical and System Research Laboratories) in Taiwan.

Reference:

1. Hung-Hsiu Yu, Kuo-Shih Tseng, Yu-Lun Ho, Mao-Feng Tu, Chun-Hung Liu, “Obstacle and cliff avoiding system and its method”.

R.O.C. Patent No: I303754 (licensed)

China Patent No. 200610082846 (pending)

U.S. Patent No: 2007/0265,740 (plea)

2. Chin-Chong Chiang, Chiu-Wang Chen, Kuo-Shih Tseng, Ta-Chih Hung, Yaw-Nan Lee, Fu-Kuang Yeh. “Apparatus with Surface Information Displaying and Interaction Capability”.

R.O.C. Patent No: I306051 (licensed)

U.S Patent No. 2008/0147,239 (pending)

Japan Patent No: 008772/2007 (plea)”.

China Patent No: 200610170233.1 (plea)

3. Kuo-Shih Tseng, Chiu-Wan Chen, Yi-Ming Chu, Hung-Hsiu Yu, Wei-Han Wang. “Method and Device of Human Robot Interaction using Tactile Sensors”.

R.O.C. Patent No: 95143158 (pending)

U.S Patent No. 2008/0134,801 (pending)

Japan Patent No: 9704/2007 (pending)”.

China Patent No: 200610168683.7 (plea)

12. [Development of Simultaneous Localization and Mapping (SLAM) for Security robot]

[2006.09~2006.12, International Cooperation Project with MIT CSAIL] [supervisor: Daniela Rus]

Abstract:

The project is a international cooperation project between ITRI and MIT CSAIL. I was visiting at Distributed Robotics Lab of MIT CSAIL for learning SLAM technology. During four months, I learned probabilistic robotics and MIT's attitude about research.

*This work is supported by Massachusetts Institute of Technology (Computer Science and Artificial Intelligence Laboratory) in U.S.A. and ITRI, Taiwan.

13. [Development of Differential Mobile Platform] [2005.1~2005.12, ITRI]

Abstract:

Recently, service robot is viewed as a potential product. Therefore, we developed a differential mobile platform for service robot.

Contributions:

I am a core member of this project. I designed the mechanical components, electrical control system and firmware of micro-controller 8051.

*This work is supported by Industrial Technology Research Institute (Mechanical and System Research Laboratories) in Taiwan.

Reference:

1. Syh-Shiuh Yeh, Meng-Chun Chen, Kuo-Shih Tseng, Mao-Feng Tu, Yang-Hua Lin “Apparatus for multi-joint lower limb exercise”.

R.O.C. Patent No: I295183 (licensed)

U.S Patent No. 2007/0142,189 (pending)

Japan Patent No: I303169 (licensed)

China Patent No: 2007/0142,190 (pending)

14. [Development of Omni-Direction Mobile Platform] [2005.1~2005.12, ITRI]

Abstract:

Recently, service robot is viewed as a potential product. Because omni-direction mobile platform is more movable than differential mobile platform. Therefore, we develop an omni-direction mobile platform for multiple applications.

Contributions:

In this project, I design the mechanical system and electrical control system and firmware of DSP are designed by NCTU students..

*This work is supported by Industrial Technology Research Institute (Mechanical and System Research Laboratories) in Taiwan.

Reference:

1. Syh-Shiuh Yeh, Meng-Chun Chen, Kuo-Shih Tseng, Mao-Feng Tu,“Method and Device for Sensing Movement of Mobile Robots”.

R.O.C. Patent No: I290881 (Licensed)

U.S Patent No: ZL 2005 1 0135278.0 (Licensed)

China Patent No: 2007/0150,096 (pending)

15. [Design and Control of Quadruped Robot Using Integrated Analog and Digital Circuits] [2002.7~2004.10, NTU]

Abstract:

The project goal is to design and implement quadruped robots which can walk on rough terrain. The detailed link is here.

The projects are supported by NTU, ITRI and MIT CSAIL.