org.usfirst.frc.team1736.lib.SignalMath

Class IntegralCalculator

java.lang.Object

org.usfirst.frc.team1736.lib.SignalMath.IntegralCalculator

public class IntegralCalculator
extends java.lang.Object

Class which implements various calculations of the integral of a signal over time. Useful for aproximating distance traveled from an accelerometer, or pose angle from a gyroscope, or in an I term in a PID, or anything else really!

The integral of a signal is pretty much just the sum of all previous values. There is some scaling dependant on how far apart those previous values were. With digital signals, it gets a bit more interesting because there's different things you can assume about what the value of the signal does between your known sample points. Often a rectangular or trapezoidal fit works great, although for lower sample rates other methods can get you further. this class implements many of these.

For the math-savvy, Wikipedia has a decent description of the various numerical integration methods this class implements for real-time.

Field Summary

Fields

Modifier and Type	Field and Description
double	accumulator Present value of the accumulator.

Constructor Summary

Constructors

Constructor and Description
IntegralCalculator(int choice_in) Initalizes everything needed for an integral calculation.

Method Summary

Modifier and Type	Method and Description
double	calcIntegral(double in) Perform numerical integration by accounting for a new sample, and returning the present integral value.
void	resetIntegral() Resets the present value of the integral to zero.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

accumulator

public double accumulator

Present value of the accumulator. Use this if you want to know the value without adding another sample to the buffer.

Constructor Detail

IntegralCalculator

public IntegralCalculator(int choice_in)

Initalizes everything needed for an integral calculation. This is the stage where you, the user, must select how to do the integration. There are 5 methods available (named 0-4). As the number increases, so does the accuracy. However, so does the time- and memory-complexity of the calculation. Choose the lowest number that suits your appliaction needs.



Integration Methods Available:

0: Rectangular rule Assumes the signal maintains the exact same value from the previous sample to this sample, and then has a step change on the new sample. "Sample and hold" is often used to refer to this assumption.

1: Trapezoidal rule Assumes the signal linearlly ramps between sample points.

2: Simpson's rule Assumes the signal acts like a 2nd-degree polynomial between points. Calculations are done by integrating a second degree polynomial splined between the 3 most recent samples.

3: Simpson's 3/8ths rule Same as Simpson's rule, but presumes a 3rd degree polynomial.

4: Boole's rule Same as Simpson's rule, but presumes a 4th degree polynomial



If you need more accuracy than this, you may wish to consider a faster sample rate.

Parameters:

choice_in - An integer between 0 and 4 inclusive to indicate the type of integration to do.

Method Detail

calcIntegral

public double calcIntegral(double in)

Perform numerical integration by accounting for a new sample, and returning the present integral value.

Parameters:

in - New sample from this control loop to add to the integral calculation

Returns:

The present value of the integral

resetIntegral

public void resetIntegral()

Resets the present value of the integral to zero. Should be called during times when the value of the integral is known to be zero. For example, if the integral is tracking robot pose angle, this could be called when the robot is backed up against a wall, since the angle is known.