FTC · For BIOBUZZ · Open source

BIOBUZZ
AutoTool.

Draw an autonomous route on the field, see what the robot will do and when, edit it by typing a request, and get compile-checked robot code.

284
Test cases
100%
Rubric score, 730 scored junctions
0
Numbers made up
2%
Timing difference from the simulator
Area
FTC
Built for
FTC Team 19859, for BIOBUZZ
Role
We designed and coded the whole system.
Status
Open source
Stack
TypeScript · Svelte 5 · Vite · Node · LLM API · javac
Credit
Built on a fork of the open-source Pedro Pathing visualizer.
What it is

What it
is.

Planning the 30-second autonomous routine usually means editing robot code by hand and finding the mistakes on the field.

AutoTool lets us draw the route on a picture of the field, see exactly what the robot will do and when, and get robot code that is checked before it is written. It is open source.

How it works

How it
works.

Draw the route

Drag the robot's stops on the field, and a time bar shows when each action happens and how many points it scores.

BIOBUZZ AutoTool: the sample blue route drawn on the field from the start to the park, the list of legs with their actions and times on the right, and the time bar along the bottom.
The sample route read straight from the robot code, each leg placed on the time bar and the AUTO points counted.

Ask in plain words

Type something like "move shot 2 six inches closer" and the change is drawn on the field for you to approve. The AI only reads the request; every number comes from code and is checked.

Check against your partner

Run our route beside our alliance partner's to catch collisions before the match, then print a plan for the drive team.

The result

What it
showed.

In testing it made up no numbers at all, and its timing came within 2% of the simulator. A full test with the AI switched on, and a timing test with students, are still to come.

What I learned

What I
learned.

Let the AI read, let code do the numbers

A language model is good at understanding a request like "move shot 2 closer", but it should not be trusted with the numbers. Keeping every number in checked code is why the tool made up none in testing.

Check before you write

Compiling the code against the real robot project before saving it means mistakes are caught on a laptop, not discovered on the field.

Plan with your partner, not just for yourself

An autonomous route that works alone can still crash into the alliance partner. Running both routes together showed that a good plan has to include the other robot on the field.