Master Thesis on Biped Robot Control Programs

November 30, 2001

Here’s the article written from my master thesis about biped control program under supervision of Prof.Prabhas Chongstitvatana, during my degree in computer engineering at Chulalongkorn University, Bangkok.

Abstract: This work presents an automatic method to synthesis robot control programs. The proposed method is based on Evolutionary Computation. The problem of biped robot walking is chosen to test the proposed method. Walking motion is divided into six stages and the evolution of control programs is carried out stage-by-stage. The locomotion is restricted to forward walking on the flat and smooth surface with static balance. The synthesis process consists of both simulation and the experiment with a real robot. The result of the experiment shows that the biped walking is achievable and stable.

C.Chaisukkosol and P. Chongstitvatana) Evolving Control Programs for a Biped Static Walker”, IEEE International Conference on Humanoid Robots, (poster paper), Waseda, Tokyo, November 22-24, 2001. [PDF: 5 pages]

>>> Read full master thesis (in Thai, with English abstract) [PDF: 69 pages A4]

>>> See master thesis presentation [PDF: 16 slides]

See video of my self-constructed wooden biped robot in experiments during developing the control program.


Bachelar thesis on Quadruped Robot Control Program

May 12, 1999

Below ‘s the article written from my bachelor thesis (with two friends) about quadruped robot control program under supervision of Dr.Jumpol Polvichai, during my degree in computer engineering at King Mongkut University of Technology Thonburi (KMUTT), Bangkok.

B. Kaewkamnerdpong, C. Chaisukkosol, O. Chanphen, and J. Polvichai, Solving the Simple Task of a Behavior-based Robot by Genetic Programming” in the First Asian Symposium on Industrial Automation and Robotics (ASIAR), BITEC, Bangkok, Thailand. May 6-7, 1999

>>> see presentation (by my friend, B.Kaewkamnerdpong) [PDF: 11 slides]

Abstract (from full Bachelor thesis)

Both Behavior-based Approach and Genetic Programming, an evolutionary computation process, are the famous successful approaches in robotics field. This work is an intermediate level approach between coding a behavior-based system and evolving an overall control system to solve the problems automatically.

In this work, the behavior-based approach has been adopted in designing a dog-like robot, which can perform behaviors such as walking forward, walking backward, turning left, turning right and stopping. To manifest the efficiency of this robot, finding the way out of the simple maze is assigned as the simple specific task of this robot. Genetic programming process is also chosen to find the programs that can command the robot out of the maze. The main evolutionary process of this work is similar to typical evolution approach. But it is adjusted especially for this work. This process is taken in the simulation. In final, the successful programs will be tested in the real robot and in the real world.

In the 4 experiments of different environments, the robot have showed that it can solve the problem well in simulation, but its efficiency decreases in the real world due to differences between simulation runs and real runs. However, this work illustrates that combination of behavior-based approach and evolutionary computation concept can solve a simple problem.

Key words :   Behavior-based Approach / Genetic Programming / Robot.

>>> see full bachelor thesis [in Thai, PDF: 114 pages A4]

Our self-constructed quadruped robot, with aluminum-legs & Tupperware-body.

The real-word experiment (along with computer simulation experiments).

My friend was picking up our quadruped robot to restart the experiment.

The first version of the quadruped robot’s legs, using foam.

%d bloggers like this: