Bevy Navigation Skill for Cursor

Bevy navigation

When to use

Load this when agents need a path over walkable space before adding a crate. Movement of the actor after a path exists belongs on bevy-character-movement. Policy for chase, patrol, and retarget belongs on bevy-ai-navigation.

ECS shape

• Components store the goal cell and the waypoint list for one agent.
• Resources own the walk grid: width, height, and blocked cells.
• Systems search on Update, then step along the path on FixedUpdate. .chain() keeps the new path visible before motion.
• Schedule: search and follow stay on the app schedules above. Crowd policy stays on bevy-ai-navigation.

Ownership

| Need | Bevy shape |

| --- | --- |

| Agent | An entity with a goal and a waypoint list |

| Walkable space | A WalkGrid resource of blocked cells |

| Avoidance | Separation stays on bevy-ai-navigation |

| Next waypoint | AgentPath stepped on FixedUpdate |

Scope

• A grid, a breadth-first search, and a follower are enough for walkable cells. Navmesh baking and crowd radius stay outside this skill.
• Actor locomotion after a waypoint exists belongs on bevy-character-movement.

Examples

[package]
name = "navigation"
version = "0.1.0"
edition = "2021"

[dependencies]
bevy = "0.20"

The grid is a resource. Search writes waypoints. The follower consumes them on the fixed step.

use std::collections::VecDeque;

#[derive(Resource)]
struct WalkGrid {
    width: i32,
    height: i32,
    cell: f32,
    blocked: Vec<bool>,
}

#[derive(Component)]
struct GoalCell(i32, i32);

#[derive(Component)]
struct AgentPath {
    waypoints: Vec<Vec3>,
    index: usize,
}

fn cell_index(grid: &WalkGrid, x: i32, y: i32) -> Option<usize> {
    if x < 0 || y < 0 || x >= grid.width || y >= grid.height {
        return None;
    }
    Some((y * grid.width + x) as usize)
}

fn search(grid: &WalkGrid, start: (i32, i32), goal: (i32, i32)) -> Vec<Vec3> {
    let Some(start_i) = cell_index(grid, start.0, start.1) else {
        return Vec::new();
    };
    if grid.blocked[start_i] {
        return Vec::new();
    }
    let mut prev: Vec<Option<(i32, i32)>> = vec![None; grid.blocked.len()];
    let mut queue = VecDeque::new();
    prev[start_i] = Some(start);
    queue.push_back(start);
    while let Some((x, y)) = queue.pop_front() {
        if (x, y) == goal {
            break;
        }
        for (nx, ny) in [(x + 1, y), (x - 1, y), (x, y + 1), (x, y - 1)] {
            let Some(i) = cell_index(grid, nx, ny) else {
                continue;
            };
            if grid.blocked[i] || prev[i].is_some() {
                continue;
            }
            prev[i] = Some((x, y));
            queue.push_back((nx, ny));
        }
    }
    let Some(_) = cell_index(grid, goal.0, goal.1).filter(|i| prev[*i].is_some()) else {
        return Vec::new();
    };
    let mut cells = Vec::new();
    let mut cursor = goal;
    cells.push(cursor);
    while cursor != start {
        let Some(i) = cell_index(grid, cursor.0, cursor.1) else {
            break;
        };
        let Some(parent) = prev[i] else {
            break;
        };
        if parent == cursor {
            break;
        }
        cursor = parent;
        cells.push(cursor);
    }
    cells.reverse();
    cells
        .into_iter()
        .map(|(x, y)| Vec3::new(x as f32 * grid.cell, 0.0, y as f32 * grid.cell))
        .collect()
}

fn plan_paths(
    grid: Res<WalkGrid>,
    mut agents: Query<(&Transform, Ref<GoalCell>, &mut AgentPath)>,
) {
    let grid_changed = grid.is_changed();
    for (tf, goal, mut path) in &mut agents {
        if !grid_changed && !goal.is_changed() {
            continue;
        }
        let start = (
            (tf.translation.x / grid.cell).floor() as i32,
            (tf.translation.z / grid.cell).floor() as i32,
        );
        path.waypoints = search(&grid, start, (goal.0, goal.1));
        path.index = 0;
    }
}

fn follow_paths(time: Res<Time<Fixed>>, mut agents: Query<(&mut Transform, &mut AgentPath)>) {
    let step = 4.0 * time.delta_secs();
    for (mut tf, mut path) in &mut agents {
        let Some(goal) = path.waypoints.get(path.index).copied() else {
            continue;
        };
        let delta = goal - tf.translation;
        let dist = delta.length();
        if dist <= step || dist == 0.0 {
            tf.translation = goal;
            path.index += 1;
        } else {
            tf.translation += delta / dist * step;
        }
    }
}

pub struct NavigationPlugin;

impl Plugin for NavigationPlugin {
    fn build(&self, app: &mut App) {
        let width = 8;
        let height = 8;
        app.insert_resource(WalkGrid {
            width,
            height,
            cell: 1.0,
            blocked: vec![false; (width * height) as usize],
        })
        .add_systems(Update, plan_paths)
        .add_systems(FixedUpdate, follow_paths);
    }
}

Mark a cell blocked and give an agent a goal. Planning runs again because the grid resource changed.

#[derive(Resource)]
struct WalkGrid {
    width: i32,
    height: i32,
    cell: f32,
    blocked: Vec<bool>,
}

#[derive(Component)]
struct GoalCell(i32, i32);

#[derive(Component)]
struct AgentPath {
    waypoints: Vec<Vec3>,
    index: usize,
}

fn block_cell(mut grid: ResMut<WalkGrid>) {
    let x = 2;
    let y = 3;
    let i = (y * grid.width + x) as usize;
    if let Some(cell) = grid.blocked.get_mut(i) {
        *cell = true;
    }
}

fn spawn_agent(mut commands: Commands) {
    commands.spawn((
        Transform::from_xyz(0.0, 0.0, 0.0),
        GoalCell(5, 5),
        AgentPath {
            waypoints: Vec::new(),
            index: 0,
        },
    ));
}

Expert Knowledge Delta

Target: Bevy 0.20+.

Core ECS types: Component, Resource, Query, Commands, Startup, Update, Fixed / FixedMain (bevy::time), .chain(), Plugin, DefaultPlugins, MinimalPlugins. Messages and observers are an ownership slot. See bevy-events-messaging for their APIs. Systems run in parallel unless .chain() orders the ones that conflict.

Who owns state

| Piece | Owner | Who may mutate | Schedule |

| --- | --- | --- | --- |

| Data on one entity | Component | One system via Query | Update or FixedMain |

| One value for the whole app | Resource | One system | Startup to insert, then the schedule that reads it |

| Spawn | Commands | The spawning system | Startup for the initial world; later schedules for runtime spawn |

| Order that must not race | .chain() | The chained systems only | Same schedule |

| Plugin boundary | Plugin | build registers systems and resources | Before the app runs |

Borrow rule: two systems that both mutably query the same component must be ordered (.chain()) or split so only one writes. Unrelated systems stay parallel. That is the ECS page rule, not a guess at a system-set API.

Decision

| If you need | Do this | Do not |

| --- | --- | --- |

| Per-entity data | Component + Query | A field stored outside a component |

| App-wide config or clock | Resource | Copy the value onto every entity |

| Stable simulation step | FixedMain and the Fixed clock | Tie simulation to raw frame Update |

| Startup spawn | Commands in Startup | Spawn from a render system |

| Cross-system notice that is not shared state | Message or observer slot in bevy-events-messaging | A new event type invented in this file |

| A path over cells | WalkGrid plus breadth-first search | An engine navmesh type this crate does not ship |

Load references/ecs-ownership.md for the ownership table.

NEVER

• Do not pin /0.20.0/ in docs URLs. WHY: links stay on latest; the 0.20 target is stated once in prose.
• Do not mutably query the same component from two unordered systems. WHY: the schedule runs systems in parallel unless .chain() orders them.
• Do not invent a physics, navigation, networking, or XR crate here. WHY: this skill only names documented Bevy crates and modules.

• No navigation types from another engine. WHY: Bevy code stays idiomatic ECS and uses only documented Bevy APIs.
• No named third-party navigation crate as the chosen solver. WHY: Bevy code stays idiomatic ECS and uses only documented Bevy APIs.
• No invented bake, avoidance, or path-query APIs. WHY: Bevy code stays idiomatic ECS and uses only documented Bevy APIs.

Reference

Official links stay on bevy.org. The library pin on bevy-master stays 0.20. No navigation module was opened.

Official Documentation

• Assets - Official asset catalog. No navigation plugin was selected.
• Examples - Example tree on the latest branch.
• ECS - Component, Query, Commands, and .chain().
• Quick Start: plugins - add_plugins without a chosen nav crate.

Related Skills

Prerequisites

• bevy-master - ECS ownership.
• bevy-character-movement - How an actor moves once a waypoint exists.

Complements

• bevy-ai-navigation - Chase, patrol, and retarget policy. Not the mesh.

Downstream / consumers

• Genre skills route chase, patrol, and similar policies.

Master

• bevy-master - Bevy router.