APM_Control: Use deceleration for stopping distance

[stephendade]

For Rover, a correction for calculating the stopping distance - it should use deceleration, not acceleration.

Tested in SITL.

[stephendade]

Turns out there's a bunch of other parts of Rover that just use acceleration instead of deceleration. Will add them to this PR...

[stephendade]

Digging further into this, the s-curve navigation assumes that the max acceleration and deceleration are the same value.

There's some fairly scary-looking maths in the s-curve navigation that I don't think I'd be able to extend to different acceleration/deceleration values.

So I'll keep this patch to the current fixing of the APM_Control portion of the code. Any flight mode that uses s-curves will continue to use the minimum of (atc_accel_max, atc_decel_max).

Tested in SITL.

[peterbarker]

This looks right, but @lthall might have words to add.

@stephendade can you explain the reasoning behind the speed_prev change, please? It's not covered in your comments.

[stephendade]

@stephendade can you explain the reasoning behind the speed_prev change, please? It's not covered in your comments.

It appeared to be a (deliberate?) typo/bug. The original if (fabsf(desired_speed) < fabsf(desired_speed) && is_positive(_throttle_decel_max)) would always return false, thus limiting the max speed change to _throttle_accel_max.

I fixed the statement to if (fabsf(desired_speed) < fabsf(speed_prev) && is_positive(_throttle_decel_max)), so the correct value of (_throttle_decel_max, _throttle_accel_max) would be used.