Deceleration curve for approaching target distance?

[kevlam2706]

We've been playing with different ways to solve the problem of "drive smoothly and quickly to a distance X away from the wall".

What about applying a simple deceleration factor with a curve?

Consider a DriveWithDecelerationCurve command:

The operator is driving at a set speed forward (e.g. 70%).

inputSpeed = 0.7

Measure the distance from the wall. Compare it to some threshold we can set (say, 1 meter). If we're nowhere near the wall, allow the motors to go at the full commanded speed.

If distance > Threshold outputSpeed = inputSpeed

If the distance is less than the threshold then start comparing the distance to the target and scale the input speed using a curve. For example:

outputSpeed = inputSpeed * (1 / distance)

or

outputSpeed = inputSpeed * (1 - (1 / distance) )

So the closer and closer we get to the target distance the more and more we slow down the target speed.

[kevlam2706]

[kevlam2706]

A graph of 1/x looks like the curve labeled "acceleration".

A graph of 1-1/x looks like the curve labeled "deceleration".

[cnnradams]

Cool I'll look into it tomorrow

On Feb 7, 2017 10:54 AM, "Kevin Lam" notifications@github.com wrote:

A graph of 1/x looks like the curve labeled "acceleration".

A graph of 1-1/x looks like the curve labeled "deceleration".

— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/FRC2706/2017-2706-Robot-Code/issues/27#issuecomment-278042553, or mute the thread https://github.com/notifications/unsubscribe-auth/AKAIuNh2LY-8l2JjRitwvbNmYi4NVuTIks5raJPRgaJpZM4L5q2x .

[ryanlarkin]

@Stealth2994, do you think that it would work if did something similar to StraightenWithDistanceSensor?

double offset = Math.abs(Robot.driveTrain.getLeftDistanceToObstacle()
    - Robot.driveTrain.getRightDistanceToObstacle());

double turnSpeed =
    baseSpeed + (offset / speedIncrease) - (speedDecrease * currentStopCycles);

[cnnradams]

The currentStopCycles thing is pretty hacky so I'd prefer to use actual math equations to do it, while using an average of the two sensors to get the distance.

[ryanlarkin]

This is pretty similar to what you did with it, maybe it could work?

Init:

double initialDistanceToObstacle = Math.abs(targetReading - currentReading);
a = (speed - minimumSpeed) / (1 - (1 / (1 + initialDistanceToObstacle)));

Execute:

double distanceToObstacle = Math.abs(targetReading - currentReading);
double driveSpeed = a * (1 - (1 / (1 + distanceToObstacle))) + minumumSpeed;

// Drive backwards in same pattern if target is behind robot
if((targetReading - currentReading) < 0)
    driveSpeed = -driveSpeed;

[cnnradams]

Yeah maybe as long as it doesn't use currentStopCycles