www.pudn.com > truetime-1.2-compiled.zip > README

----------------------------------------------
 TrueTime version 1.2
 Example: Task Scheduling and Control
 Directory: $DIR/examples/threeservos
----------------------------------------------

1. INTRODUCTION

This example extends the simple PID control example (located in
$DIR/examples/simple_pid) to the case of three PID-tasks running
concurrently on the same CPU controlling three different servo
systems. The effect of the scheduling policy on the global control
performance is demonstrated.


2. COMPILING THE SIMULATION FILES

2.1 Configuring the TrueTime Environment

    Before starting Matlab, you must set the environment variable
    TTKERNEL to point to the directory with the TrueTime kernel files:

        Unix/Linux: > export TTKERNEL=$DIR/kernel
        Windows:  Control Panel / System / Advanced / Environment Variables

    Then add the following lines to your MATLAB startup script. This
    will set up all necessary paths to the TrueTime kernel files.

        addpath(getenv('TTKERNEL'))
        init_truetime;

2.2 Compilation

    As described in the reference manual it is possible to write a
    TrueTime simulation (i.e. the code functions for the tasks and the
    initialization commands) either as m-files or as C++ functions.
    Both approaches are provided for this example and the compilations
    needed for the respective alternatives are described below.

    2.2.1 Matlab version
   
	TrueTime in the MATLAB version is compiled once and for all 
	by running the command make_truetime from the command prompt.
	This should be done the first time you are using TrueTime and
	no further compilation is then required.

    2.2.2 C++ version
    
        Compilation of the init script of the computer block is
        performed by the command:

        >> ttmex threeservos_init.cpp
    
	NOTE: If changes are made to the code function or the
        initialization commands, the init-file needs to be re-compiled.


3. SIMULATIONS

Open the model threeservos.mdl to run the simulation.

- Make sure that rate-monotonic scheduling is specified in the
  function ttInitKernel and simulate the system. Study the computer
  schedule and the control performance. Task 1 will miss all its
  deadlines and the corresponding control loop is unstable.

- Change the scheduling policy to earliest-deadline-first and run a
  new simulation. Again study the computer schedule and the control
  performance. After an initial transient all tasks will miss their
  deadlines, but the overall control performance, however, is
  satisfactory.