Gyro Turns and Lag

The mission 4 min

Turn to the angle you actually asked for — and find out why the robot keeps going after you tell it to stop.

Open lesson 14 in your journal and find the three turn measurements: 90, 45 and 180, asked for against achieved. Those are the numbers today beats. Write them at the top of a fresh page before you start.

The corners you have to hit 5 min

Work in the reef square. A to B across the north edge, then the turn down towards D, then west along the south edge to C. Three legs, three corners.

These are the turns that decide the block. A few degrees lost at the first corner puts the robot off-centre at the second, and by the third the attachment arrives somewhere the item is not — which is the error accumulation you met in lesson 15, now with real geometry behind it.

The technique — why equals fails 12 min

The obvious program is: start turning, wait until the gyro reads 90, stop. It fails, and it fails in two different ways that are worth separating.

First, the loop can miss the number entirely. Your program checks the angle, does a little work, checks again. Between two checks the robot has turned by more than one degree. If it was at 89 and the next reading is 91, the angle was never equal to 90 — and a program waiting for equality waits forever while the robot spins.

The fix is to stop asking whether the angle equals the target and start asking whether it has changed by more than the target. Once it has passed the value, it stays passed, and the test cannot be missed no matter how coarse the loop.

Second, the robot does not stop when you say stop. Even with a test that fires reliably, the wheels are turning when the decision is made, the motors take a moment to brake, and the robot carries on a few degrees past. That is lag, and no amount of better testing removes it — it is momentum.

Today’s answer to lag is deliberately crude: aim off. If the robot consistently overshoots by five degrees, ask for five fewer. It feels like cheating and it is entirely standard — you are correcting for a known, repeatable physical effect. The elegant answer, slowing down as the target approaches, is proportional control, and it arrives next lesson.

Build it — decide the pivot 10 min

No building, one decision, written in the journal because everything after it depends on the answer.

Point turn or pivot turn? A point turn spins both wheels in opposite directions and rotates the robot about its own centre. A pivot turn stops one wheel and swings around it.

Point turns keep the robot where it is, which matters when it is already lined up on a station. Pivot turns move the robot as they turn it, which is sometimes exactly what you want between reefs, but means the turn and the travel are tangled together.

Pick one for this block, use it consistently, and note which. A team that mixes both will find that a turn which worked at one reef behaves differently at the next.

Program it — rebuild Turn By Degrees 12 min

You already have a Turn By Degrees My Block from lesson 14, driven by wheel rotations. Today its insides are replaced while its name and its input stay exactly the same.

That is the reward for lesson 13 arriving before this one. Every program that calls it inherits the improvement without being touched.

Inside the block:

• reset the gyro angle to zero
• start the motors turning in the chosen direction
• wait until the angle has changed by more than the requested amount
• stop the motors

Keep the input meaning what it meant before, including which direction a positive number turns — you wrote that down in lesson 14 and every existing route depends on it.

Tune it — measure the overshoot 9 min

Ask for 90 from a marked heading, three times, and measure what you actually get each time. Then 45. Then 180.

You are looking for two things. How big is the overshoot, and is it consistent? A consistent overshoot is a gift: subtract it from the request and the turns come good. An inconsistent one means something is loose, the battery is low, or the gyro is not rigidly mounted.

Once you have aimed off, run all three again and put the new numbers beside lesson 14’s. That table is the point of the lesson.

Run it — three corners 6 min

Drive the reef square: A to B, turn; B to D, turn; D to C. Use your rebuilt Turn By Degrees for every corner.

Finish by checking the robot’s heading against where it should be. Three turns is enough to reveal whether your correction holds up when errors have a chance to stack — and if it does not, you have found the thing lesson 23 will have to solve.

The journal 4 min

Today’s entry:

• point turn or pivot turn, and why
• the before-and-after table: lesson 14’s three measurements beside today’s, for 90, 45 and 180
• your overshoot at each angle, and whether it was consistent
• the amount you aimed off by
• one sentence on why waiting for the angle to equal a value fails — this is the bug most likely to come back, and writing it down is how you recognise it next time
• the heading error after three corners