Task System

A task is a small, composable unit of robot behavior: it is set up once, then polled every loop until it reports that it is done. Chain and nest tasks with fluent builders to describe an entire autonomous routine — or a single asynchronous action in TeleOp — declaratively.

The task model

Every action extends the abstract Task class, which has exactly two lifecycle methods:

• void init() — called once, before the first run(). Set up state and issue starting commands here (for example, tell a follower to begin a path).
• boolean run() — called every loop after init(). Return true when the task is finished, or false to keep running.

A task never blocks. Instead of looping internally, it does a tiny slice of work each run() and returns a "finished yet?" flag, so the OpMode loop stays responsive and can run many things at once.

TaskMaster — running one task at a time

A TaskMaster drives a single active task to completion, one loop at a time. It starts idle (holding a NullTask), and you feed it work with addTask(Task):

• If the master is idle, the new task becomes active and is initialized on the next update().
• If the master is busy, the new task is appended after the current one — the master wraps them into a SequentialTask, so calls to addTask() queue up in order.

Call update() exactly once per loop. It initializes the active task on the first call, runs one step, and returns to idle once the task finishes. getStatus() reports whether the active task finished on the most recent update, and getTask() returns the currently active task (a NullTask when idle).

TaskMaster master = new TaskMaster();

// every loop:
master.addTask(new SleepTask(500));   // queues; runs after whatever is already active
master.update();                      // advance the active task one step

Composing tasks

Because run() is just "am I done yet?", tasks nest cleanly. The Task base class exposes four fluent builders, each returning a new composite Task:

Task routine =
    new InstantTask(() -> arm.raise())
        .andThen(new SleepTask(300))
        .alongWith(new BasicTask(intake::start, intake::hasSample))
        .raceWith(new SleepTask(2000));   // bail out after 2s no matter what

master.addTask(routine);

The builders take varargs and return a Task, so you can keep chaining. Under the hood andThen(a, b) is simply new SequentialTask(this, a, b) — you can always build the composites directly if you prefer.

Task catalog

The "finishes when" column reflects each class's real run() implementation.

PedroPathing tasks

Two tasks bridge to a PedroPathing Follower, letting you drop path following straight into a task chain.

Pedro 3.0 FollowPathTask

Follows a com.pedropathing.paths.Path. In init() it calls follower.follow(path); in run() it returns true once follower.isBusy() is false. An optional third constructor argument is a speed multiplier (a fraction of the robot's max speed, 1.0 = full speed):

new FollowPathTask(follower, path);          // full speed
new FollowPathTask(follower, path, 0.6);     // 60% of max speed

Pedro 3 dropped the per-follow speed argument of the old followPath() call. A speed cap is now a path modifier — Foresight.config.maxPathSpeed, a fraction of the robot's max speed — so when speed < 1.0 and the follower's algorithm is a Foresight, the task applies the cap to a copy of the path via path.with(config.maxPathSpeed.at(speed)). Only Foresight exposes this knob; with other algorithms the path is followed at full speed.

Pedro 3.0 HoldPointTask

Commands the follower to actively hold a com.pedropathing.math.Pose by calling follower.hold(pose) in init(), then finishes immediately. Chain it after a path to lock the robot in place before the next action begins.

An autonomous chain

Path toBasket = /* build a com.pedropathing.paths.Path */;
Path park     = /* build a com.pedropathing.paths.Path */;

Task auto =
    new FollowPathTask(follower, toBasket)
        .andThen(new InstantTask(() -> scorer.deposit()))
        .andThen(new SleepTask(400))
        .andThen(new FollowPathTask(follower, park, 0.75))
        .andThen(new HoldPointTask(follower, park.endPose()));

master.addTask(auto);