CS452 - Real-Time Programming - Spring 2009

Lecture 24 - Calibration

Controlling a Train

Types of Control

Ballistic Control

Feedback Control

Summary

The sensors give us large granularity feedback control.

Between the sensors we must use ballistic control,

• which depends on integrating the train speed over time.

Calibration

Based on

• using knowledge of the past to predict the future

Methodology

• Measure the train's speed
• Assume that the future will be like the past
• Then the speeed of the train under the same condition will be the same in the future

Measuring Speed

The easy part

• Get time at sensor N
• Get time at sensor N+1
• Calculate speed between the sensors using
  • distance between sensors
  • difference between times

Collect many such measurements

Using a Calibration

The other easy part

1. Arrive at sensor N at time t
   • Use clock server to label sensor event with time
2. Look up velocity to use in calibration table
3. Position at time t' is sensor position + velocity x (t' - t)

Building a Calibration Table

The hard part.

Two things are hard.

1. Dealing with measurement error
   • Two categories of error
     1. Systematic = controllable

        Examples,

        • measurement lag
          • Assume average lag is always the same
          • Calculating time intervals works fine
          • Calculating actual position at a specific time needs a lag added
          • How do you find out the lag?

          curved versus straight

        Solution

        • Subdivide the data
        • But not too much
     2. Random = uncontrollable

        Examples,

        • variations in measurement lag

        Level of random error determines ultimate accuracy of calibration

        Solution

        • Take more measurements
        • But without consuming much time

        Rule of thumb

        • Many measurements, simple algorithm works better than
        • Few measurements, complex algorithm
2. Determining what state is relevant

   State is has many potentially relevant aspects

   • train speed
   • which train it is
   • curved versus straight versus switch
   • previous speed
   • time since last speed change
   • cleanliness of track
   • time since last maintenance of train

Practical Issues

1. You are already doing a whole lot of measurements
2. Average in a circular buffer
3. Size & align table by size & alignment of cache lines
   • linker command script

Return to:

• Bill Cowan's lecture notes for CS452 in Spring 2009
• Bill Cowan's Spring 2009 CS452 page
• Bill Cowan's CS452 page
• Bill Cowan's teaching page
• Bill Cowan's home page