bevy_rl

Build Reinforcement Learning Gym environments with Bevy engine to train AI agents that learn from raw screen pixels.

Compatibility

| bevy version | bevy_rl version | | ------------ | :-------------: | | 0.7 | 0.0.5 | | 0.8 | 0.8 |

Features

• Set of APIs to implement OpenAI Gym interface
• REST API to control an agent
• Rendering to RAM membuffer

Usage

1. Define Action Space and Application State

```rust

[derive(Debug, Clone, Eq, PartialEq, Hash)]

enum AppState { InGame, // Actve state Control, // A paused state in which application waits for agent input Reset, // A request to reset environment state }

// List of possible agent actions (discrete variant) bitflags! { #[derive(Default)] pub struct PlayerActionFlags: u32 { const IDLE = 1 << 0; const FORWARD = 1 << 1; const BACKWARD = 1 << 2; const LEFT = 1 << 3; const RIGHT = 1 << 4; const TURNLEFT = 1 << 5; const TURNRIGHT = 1 << 6; const SHOOT = 1 << 7; } } ```

2. Enable AI Gym Plugin

```rust let gymsettings = AIGymSettings { width: 256, height: 256, numagents: 2, };

app
    // bevy_rl initialization
    .insert_resource(gym_settings.clone())
    .insert_resource(Arc::new(Mutex::new(AIGymState::<PlayerActionFlags>::new(

```

3. Make sure environment is controllable at discreet time steps

rust struct DelayedControlTimer(Timer);

```rust app.insertresource(DelayedControlTimer(Timer::fromseconds(0.1, true))); // 10 Hz app.addsystemset( SystemSet::onupdate(AppState::Control) // Game Systems .withsystem(turnbasedtextcontrolsystem) // System that parses user command .withsystem(executeresetrequest), // System that performs environment state reset );

app.addsystemset( SystemSet::onupdate(AppState::InGame) .withsystem(turnbasedcontrolsystem_switch), );

```

```rust fn turnbasedcontrolsystemswitch( mut appstate: ResMut>, time: Res, mut timer: ResMut, aigymstate: ResMut>>>, ) { if timer.0.tick(time.delta()).justfinished() { appstate.push(AppState::Control); physics_time.pause();

    let ai_gym_state = ai_gym_state.lock().unwrap();
    ai_gym_state.send_step_result(true);
}

} ```

4. Handle Reset & Agent Actions from REST API in Bevy Environment

```rust pub(crate) fn executeresetrequest( mut appstate: ResMut>, aigymstate: ResMut>>>, ) { let aigymstate = aigymstate.lock().unwrap(); if !aigymstate.isreset_request() { return; }

ai_gym_state.receive_reset_request();
app_state.set(AppState::Reset).unwrap();

}

pub(crate) fn turnbasedcontrolsystemswitch( mut appstate: ResMut>, time: Res, mut timer: ResMut, aigymstate: ResMut>>>, aigymsettings: Res, mut physicstime: ResMut, ) { if timer.0.tick(time.delta()).justfinished() { appstate.overwritepush(AppState::Control).unwrap(); physics_time.pause();

    let ai_gym_state = ai_gym_state.lock().unwrap();
    let results = (0..ai_gym_settings.num_agents).map(|_| true).collect();
    ai_gym_state.send_step_result(results);
}

}

pub(crate) fn turnbasedtextcontrolsystem( agentmovementq: Query<(&mut heron::prelude::Velocity, &mut Transform, &Actor)>, collisionevents: EventReader, eventgunshot: EventWriter, aigymstate: ResMut>>>, aigymsettings: Res, mut appstate: ResMut>, mut physicstime: ResMut, ) { let mut aigymstate = aigymstate.lock().unwrap();

if !ai_gym_state.is_next_action() {
    return;
}

let unparsed_actions = ai_gym_state.receive_action_strings();
let mut actions: Vec<Option<PlayerActionFlags>> =
    (0..ai_gym_settings.num_agents).map(|_| None).collect();

for i in 0..unparsed_actions.len() {
    let unparsed_action = unparsed_actions[i].clone();
    ai_gym_state.set_reward(i, 0.0);

    if unparsed_action.is_none() {
        actions[i] = None;
        continue;
    }

    let action = match unparsed_action.unwrap().as_str() {
        "FORWARD" => Some(PlayerActionFlags::FORWARD),
        "BACKWARD" => Some(PlayerActionFlags::BACKWARD),
        "LEFT" => Some(PlayerActionFlags::LEFT),
        "RIGHT" => Some(PlayerActionFlags::RIGHT),
        "TURN_LEFT" => Some(PlayerActionFlags::TURN_LEFT),
        "TURN_RIGHT" => Some(PlayerActionFlags::TURN_RIGHT),
        "SHOOT" => Some(PlayerActionFlags::SHOOT),
        _ => None,
    };

    actions[i] = action;
}

physics_time.resume();
control_agents(actions, agent_movement_q, collision_events, event_gun_shot);

app_state.pop().unwrap();

} ```

REST API

| Method | Verb | bevy_rl version | | ----------------- | -------- | ------------------------------------------ | | Camera Pixels | GET | http://localhost:7878/screen.png | | Reset Environment | POST | http://localhost:7878/reset | | Step | POST | http://localhost:7878/step body=ACTION |