Here are some use cases where ezAuton would and wouldn't be a good idea to use
- β Creating a codebase that will run on FTC, FRC, and simulations without modifications
- β Type safetyβdesigned around units
- β Using a Java only codebase. ezAuton uses Kotlin heavily due to the multiple advantages it gives, particularly with suspending functions (approximately equivalent to async/await), and DSLs
- β Likes the idea of structured concurrency and coroutines
- β Wanting to work on robot code without a physical robot
- β Recording/streaming robot telemetry data
- β Playing back or live viewing robot telemetry data
- β Using trajectory control algorithms with minimum boilerplate
- β Programming an FTC Robot
- β Programming an FRC Robot
The following example is all that is needed for a simulation running pure pursuit that is recorded. Note this is very similar for any trajectory algorithm, I am aware pure pursuit is inferior to other methods like ramsete.
suspend fun run() {
// (1) a straight line trajectory starting at (0,0) going to (0,20) with a max speed of 3 m/s.
val trajectory = trajectory(samplePeriod = 5.ms) {
point(0.m, 0.m, speed = 0.mps, acceleration = 13.0.mps / s, deceleration = 12.0.mps / s)
point(0.m, 10.m, speed = 3.mps, acceleration = 13.0.mps / s, deceleration = 12.0.mps / s)
point(0.m, 20.m, speed = 0.mps, acceleration = 13.0.mps / s, deceleration = 12.0.mps / s)
}
// (2) a simulated robot
val robot = SimulatedTankRobot.create(lateralWheelDistance = 1.m, maxAccel = 14.0.mpss, minVel = 0.3.mps, maxVel = 16.0.mps)
// (3) a lookahead that scales with the velocity of the robot
val lookahead = ScalingLookahead(distanceRange = 1.0.m..5.0.m, speedRange = 2.0.mps..10.0.mps, velocityEstimator = robot)
// (4) pure pursuit
val purePursuit = robot.purePursuit(period = 10.ms, trajectory, lookahead)
// (4) the action we will actually be running
val action = action {
// (5) record everything inside this
val recording = recording {
// (6) include data about the path
include(trajectory.path.simpleRepr)
// (7) run pure pursuit in parallel
parallel(purePursuit)
// (8) every 10ms sample data from the robot (like location) and include in recording
sample(10.ms, robot)
}
// (9) save the recording to ~/.ezauton/test.json
recording.save("test.json")
}
action.run()
}
-
Create a trajectory.
- Note the use of extension properties like
.m, .mps, .mps/s = .mpss
. The library uses units, and they can usually be multiplied, added, or divided together as one would do in maths or physics. The values stored in each container is the SI value. If someone wanted to use feet instead of meters, they could use.ft
and likewise for other units. - Trajectories contain a path which is comprised of path segments (in this case linear path segments)
- Trajectories also contain a speed at any distance along the path
- We specify starting, stopping points, and acceleration/decelerations.
- The trajectory will attempt to optimize time for given constraints (deceleration at last moment)
- The trajectory generates sub-trajectories every 5ms which are linearly interpolated between
- Note the use of extension properties like
-
Create a simulated robot.
-
A simulated robot implements a lot of interfaces. This is useful because this library is built around interface and not implementation. Trajectory algorithms and other functions almost always accept an interface, so the library is flexible to any implementation.
-
Notice that some of the interfaces are oddly named. Generally, abbreviations are used at a minimum, but to avoid extraordinarily long names, a few are used. For example
- Rot = Rotation
- Vel = Velocity
- Trans = Translational
- Loc = Location
- Est = Estimator
-
-
Create a lookahead that scales with the speed of the robot
robot
implementsTankRobotVelEst
, an interface which in turn implementsVelocityEst
, so it can be passed into velocity estimator
-
Create a pure pursuit action.
-
This uses a useful extension function to allow for less boilerplate.
-
The extension function depends on any classes which implement both
TransLocEst
andTransLocDrivable
. This means that the robot can estimate its translational location and drive towards any given translational location. The tank robot implementation ofTransLocDrivable
implements this by driving in arcs. -
If there is no common interface, it is easy to use the regular non-extension version
-
-
Record anything inside the scope into a Recording object which can be saved
-
Include the data of the path. Currently, only a simple representation of the path which is just a list of points can be serialized. This is because paths can contain any type of path segments---including weird shapes such as curves, which might be hard to serialize and even harder to display in a visualizer
-
Run pure pursuit in parallel (so we can sample and run pure pursuit at the same time).
- Pure pursuit sees that it is in a recording scope and records data each time it loops
-
Sample data (such as location) every 10 milliseconds.
- robot implements
Sampler<Data.TankRobotState>
, so it can be sampled
- robot implements
-
Save the data to a json file located in
test.json