Introduction

In a previous work (##Val98 ##Val98), the closed architecture of the robot arm was opened by changing the robot's computer by a PC plus some data acquisition cards, like the jr3 force sensor, an encoders reader, A/D and D/A converters. All of this was done following the electrical specifications. The architecture was transformed so the control algorithm could be changed and external sensors could be used. But still the architecture was closed since the operating system used was Windows NT. The main disadvantages of this operating system were:

$$\mbox{ \epsfxsize =70mm \epsffile{images/robot.eps}}$$

$\parbox{7cm}{{\bf FIGURE \arabic{figcounter}:}\quad {\it Robot arm. } }$

$$\mbox{ \epsfxsize =70mm \epsffile{images/robothand.eps}}$$

$\parbox{7cm}{{\bf FIGURE \arabic{figcounter}:}\quad {\it Robot arm hand. } }$

• Very poor real time performance.
• the scheduling policy was difficult to change.
• There was no way of seeing what was happening in the system, i.e. tasks activations and compute time.

So, we decided to move to RT-Linux (##Bar96 ##Bar96). Now we got completely control over the entire process from the point of view of control and real time. So, it has been built a platform where it can be tested the strong relations between real time theory and control theory. In order to show the capabilities of the system, as a control example, it has been developed an application to make the robot arm play chess. In this hard rtos, it has been implemented the control of a robot arm, and some useful tools to investigate the effects of scheduling in control applications. To prove the platform, a real time application has been developped. This application combines pattern recognition techniques (voice recogniton and image processing) and real time ones to make the robot arm play chess.