Team Description Paper for RoboCup 2004

RoboCup and FIRA are two international robotics game organizations which hold annual humanoid-robots soccer-playing competitions. Their quest: "Develop a team of humanoids to play against human World Champions by the year 2050." The aim of this project is to build small size (height being shorter than 0.5m) humanoid robots that can win these competitions. These robots are designed using off-the-shelf components which can be found easily in hobby shops. We named it ROPE, which stands for RObot PEle or RObot for Personal Entertainment, representing our desire and determination to make it play soccer with excellent skill.

The robot will have to be able to detect and approach a ball, and be able to dribble it through the opponents and deliver the ball into the goal. Different strategies and decision making algorithms will be explored. This project is also an excellent platform for technological education, especially in the areas of control, mechanical design, machine vision, system integration.

The first version ROPE-I has been built to test the performance of several components such as actuators and sensors as well as the mechanical design. We have managed to make it walk statically. A short video clip can be found here (QuickTime - 4.5MB). With enough experience and confidence gained from developing ROPE-I, we are now building another biped with stronger structure and higher height than the older. Images of ROPE-I and ROPE-II can be seen below.

ROPE-I

ROPE-II

ROPE-II

Specifications of ROPE-II

Control system

The performance of a biped directly depends on the control system. If the control system is not sufficient, no matter how good is the control algorithm, the biped may not have desired performance. However, the power of a control system is usually inversely proportional to its size. We need to built a powerful control system but within space limit of the application.

The control system used in ROPE-II consists of computer system, sensors and actuators. Followings is a diagram illustrating connection among various elements of the control system.

Computer system

• PC-104: This is really a normal Personal Computer with compact size suitable for mobile application.
• DAQ card: The card we are using is DMM32AT of Diamond System. This DAQ card is used to read analog signals from the sensors.
• Basic Stamp: This is a microcontroller which is built from popular PIC chip family. Basic Stamp receives position commands from PC-104 and generate PWM signals to control the servos
• Operating system: RTLinux is the operating system of choice. Among the advanges of using this OS are its free distribution and real-time behavior. Its performance is so great.

Sensors

There are several sensors needed for the biped to walk stably. These include:

• Force sensor: Force sensors are used to sense the center of pressure on the feet sole. We are using FlexiForce with measurement range of 25lbs.
• Tilt sensor: This dual axis tilt sensor is used to sense the tilt angles of the body. This is a product of Crossbow.
• Accelerometer: The 3-axis accelerometer in use is also a product of Crossbow. It is used to sense the acceleration of the robot.
• Gyro: This sensor of Silicon Sensing System is used to sense the direction in which the biped is heading.
• Camera: This is the vision sensor of the robot. We are using CMUCam which is a low cost yet powerful camera.

Actuators

Actuators are also a very important element of a biped. A perfect actuator should be a light-weight yet high-torque motor. We have found that servo motors, which are used extensively in hobby aeroplane, are very suitable for this application. Initially we used Futaba motor, but it is not strong enough with a stall torque of 8 kgcm. We are now using Hitec motor with torque up to 13 kgcm. That means we can hang a 13 kg object at the arm length 1cm!

Control algorithms

To have a good control system is just half of the solution. A robust conrol algorithm is the other half. Several control algorithms have been studied and simulated to find a suitable algorithm for this robot. Among the algorithms of studying are static walking, virtual model control, linear inverted pendulum, ZMP trajectory planning,... Some simulations can be found in the movie clips section. With the solid background gaining from studying existing algorithms, we will develop a new algorithm suitable for this biped.

Movie clips

Static Walking of ROPE-I [*.wmv format - 3MB] [QuickTime format - 4.5MB]

Static Walking of ROPE-II [*.wmv format - 3MB] [QuickTime format - 4.5MB]

ROPE-II kicks a ball [*.wmv format - 7MB]

Head tracks a ball [*.avi - 10MB]

Directional control of ROPE-II [*.wmv-22MB]

Static Walking Simulation [*.wmv format - 1MB] [QuickTime format - 5MB]

Linear Inverted Pendulum Simulation [*.wmv format - 2MB] [QuickTime format - 4.5MB]

Virtual Model Control using Genetic Algorithm to optimize ankle gain [*.wmv format - 2MB]

Last Updated: 17 March 2004