#trajectory #bevy #ryot #ray-cast #game-mechanics

yanked ryot_trajectories

Implements trajectory capabilities for Bevy, crucial for interactive game mechanics like line-of-sight, fog, complex collision, etc

1 unstable release

0.2.2 May 6, 2024

#12 in #ryot


Used in ryot_tiled

AGPL-3.0-only

120KB
2K SLoC

Ryot Trajectory

License Crates.io Downloads Docs Discord

What is Ryot Trajectory?

Ryot Trajectory leverages the concept of ray casting to provide a robust trajectory system for Bevy. It is designed to work seamlessly with Bevy's ECS, enabling developers to implement advanced trajectory-based mechanics in their games and simulations that require precise trajectory calculations, with support for obstacles, line of sight, and other complex interactions. Even though it's optimized for 2D grid-based environments, it can be easily extended to fit specific game requirements and open-world scenarios. It's part of Ryot framework, having Ryot Core and Ryot Utils as dependencies.

Ray Cast and Trajectories

Trajectory is a fundamental concept in game development, enabling complex game mechanics such as projectile motion, line of sight, fog of war, collision detection, and more. Ray casting is a technique used to simulate trajectories by tracing rays through a 2D or 3D environment, detecting collisions and interactions along the way. It's widely used in games to implement realistic physics, lighting effects, and AI behaviors, providing a versatile tool for creating engaging gameplay experiences.

In the context of Ryot Trajectory offers a comprehensive solution for implementing trajectory systems in Bevy projects, providing a flexible and extensible framework for handling trajectory logic. By leveraging the ECS architecture and seamless Bevy integration, developers can create dynamic trajectory systems that adapt to changing game conditions and player interactions.

Trajectory uses Bevy RayCast3d as the underlying ray casting library, for both 2D and 3D environments.

Capabilities

  • Seamless Bevy Integration: Built to work hand-in-hand with Bevy's ECS, offering smooth integration and ensuring compatibility with Bevy's event systems.
  • Ray Casting Support: Utilizes ray casting to simulate trajectories, enabling precise collision detection and interaction with obstacles.
  • 2D Optimization: Specially tailored for 2D grid-based navigation, providing robust tools for tile-based and open-world game environments.
  • Extensible Architecture: Designed to be flexible, allowing developers to extend and customize trajectory logic to fit specific game requirements.

Basic Setup

Before setting up the trajectory framework, lets understand the core concepts: Point, TrajectoryPoint, Navigable, RadialArea<P>, Perspective<P> and Trajectory<T, P>.

Point

The Point trait represents a position in the world. It's a core concept of the Ryot ecosystem, that allows you to integrate your own world representation with Ryot and its spatial algorithms.

TrajectoryPoint

An extension of the Point trait, the TrajectoryPoint trait represents a position in the world that can be used to calculate a trajectory. It's used to generate the bounding box of a given point in space, used to check the point against the ray cast. This crate uses primarily the aabb3d (axis-aligned bounding box) to represent the bounding box of a point in space and the ray cast aabb intersection to check if a point is inside a trajectory or not.

Navigable

The Navigable trait belongs to ryot_core and is used to determine if a point is navigable or not. It's used to determine if an actor can go through a particular point in the world, for instance if this point is walkable or not.

Currently, Navigable has two flags: is_walkable and is_flyable. The first one is used to determine if an actor can walk through a point, and the second one is used to determine if an actor can fly through a point.

RadialArea

The RadialArea struct is a descriptive representation of an area in the game world. It contains only primitive and copyable types, implements Hash and its main purpose is to be used as a descriptive representation of Perspectives, allowing to cache complex calculations of perspectives and reuse them in the future.

Perspective

The Perspective struct is a representation of a perspective from a given spectator point. It contains an array of traversals, which are tuples of RayCast3d and the area traversed by the ray. It's used to represent all the trajectories that a spectator can see from a given point in space in a determined scenario/condition.

Trajectory<T, P>

The trajectory struct is a representation of a trajectory in the game world. It's a component that can be attached to entities in the ECS, and it's used to represent the trajectory request of an entity. It contains the radial area of the trajectory, the conditions that the trajectory must satisfy, the entities that the trajectory can be shared with, and a set of params that can be used to customize the trajectory calculation.

Bevy

To integrate ryot_trajectories you need to add a trajectory to your Bevy app. This is done by calling the add_trajectory<T, P, N>, where T is a marker type that represents the trajectory context, P a TrajectoryPoint type and N is the Navigable type. This method is a builder method on your Bevy app builder.

Here is a basic example:

use bevy::prelude::*;
use ryot_core::prelude::*;
use ryot_trajectories::prelude::*;

fn setup<P: TrajectoryPoint + Component>(mut commands: Commands) {
    // here we use () as a marker, but in a real scenario you should use a marker type
    // that properly represents the context of the trajectory.
    commands.spawn(Trajectory::<(), P>::default());
}

fn build_app<P: TrajectoryPoint + Component>(app: &mut App) -> &mut App {
    app
        .add_plugins(DefaultPlugins)
        .add_trajectory::<(), P, Flags::default()>()
}

Components

This crate has two main ECS components:

TrajectoryRequest<T, P>

This component is attached to entities that require a trajectory computation. It specifies the parameters for the trajectory algorithm:

  • area: the radial area that represents the area that the trajectory will cover.
  • shared_with: the entities that the trajectory can be shared with.
  • conditions: the conditions that the trajectory must satisfy, based on a navigable type and a position.
  • params: a set of parameters that can be used to customize the trajectory calculation.
    • max_collisions: the maximum number of collisions that a ray cast in this trajectory can have.
    • reversed: if the trajectory should be analysed in reverse order (from the end to the start).
    • execution_type: the type of execution that the trajectory should have: once or time based.
  • last_executed_at: the last time that the trajectory was executed, a flag to determine if the trajectory should be executed again or not.

It's part of the public API and should be used by the user to trigger trajectory computations.

TrajectoryResult<T, P>

This component is attached to entities that have completed a trajectory computation. It holds the result of the trajectory computation, represented by:

  • collisions: the collisions between the trajectory and the world, meaning that the trajectory is not navigable in these points.
    • position: the position where the collision happened.
    • distance: the distance from the start of the trajectory to the collision.
    • previous_position: the previous position of the trajectory, before the collision.
    • pierced: if the collision was pierced or not, meaning that the trajectory continued after the collision.
  • area_of_interest: the area of interest of the trajectory, meaning that the trajectory is navigable and can influence the world in these points.

This component is attached to entities that have completed a trajectory computation. It's part of the public API and should be used by the user to check the trajectory results.

Systems

The trajectory framework is composed of three main systems:

  1. update_intersection_cache<T, P>: this system updates the cache of intersections represented by a radial area present in the TrajectoryRequest<T, P> component. This cache is used to speed up the trajectory computation, avoiding re-calculating already calculated ray cast aabb intersections.
  2. process_trajectories<T, P, N>: the main system of the trajectory framework, it executes the trajectory requests present in the ECS, calculating the trajectories and attaching the results to the entities.
  3. share_results<T, P>: this system shares the trajectory results of an entity with the entities that the trajectory can be shared with.

There are also two systems that are part of the clean-up process:

  1. remove_stale_results<T, P>: this system removes the trajectory results that are no longer associated to a trajectory request.
  2. remove_stale_trajectories<T, P>: this system removes the trajectory requests that are no longer valid.

Examples

Choose an example to run based on your needs, such as handling multiple entities or dealing with obstacles:

cargo run --example example_name --features stubs

Replace example_name with the name of the example you wish to run.

Understanding the Examples

Each example included in the library showcases different aspects of the trajectory system:

  • Basic: Demonstrates a basic complete trajectory use case, with obstacles and different radial areas.
  • Stress Test: Evaluates the trajectory's performance under high load conditions.

Building Your Own Scenarios

Leverage the ExampleBuilder to customize and create tailored trajectory example/test scenarios:

fn main() {
    // ExampleBuilder::<T /* Contextual Marker */, P /* TrajectoryPoint */, N /* Navigable */>::new()
    // .with_trajectories(/* array of (trajectory, count) tuples, containing the trajectories to be instantiated and how many */)
    //  .with_obstacles(/* number of obstacles to be instantiated */)
    //      .app() // basic app with visual capabilities
    //      /* add your custom systems, plugins and resources here */
    //      .run();
}

Benchmarks

Performance benchmarks are included to provide insights into the crate's efficiency. The benchmark can be run to evaluate performance under various conditions:

cargo bench --features stubs

Results

There are three main benchmarks for the trajectory system: creating trajectories, executing trajectories, and checking navigable points against trajectories. The benchmarks cover different scenarios, such as linear, sectorial and circular areas, with different ranges values.

The following tables provide an overview of the benchmark results for the trajectory system:

Creation

Test Name Type Range (Distance) Time (ns/iter) Variability (± ns) Iterations per Second (iters/s)
create_linear_range_10 linear 10 143 3 6,993,007
create_linear_range_100 linear 100 821 114 1,218,027
create_linear_range_255 linear 255 1,337 197 747,951
create_45_degrees_sector_range_10 radial_45 10 1,160 12 862,069
create_45_degrees_sector_range_100 radial_45 100 17,142 409 58,358
create_45_degrees_sector_range_255 radial_45 255 29,580 830 33,822
create_90_degrees_sector_range_10 radial_90 10 2,734 112 365,632
create_90_degrees_sector_range_100 radial_90 100 34,297 883 29,159
create_90_degrees_sector_range_255 radial_90 255 59,535 2,329 16,802
create_circular_range_3 circular 3 1,871 28 534,759
create_circular_range_5 circular 5 4,724 74 211,640
create_circular_range_10 circular 10 9,819 292 101,844
create_circular_range_25 circular 25 38,055 769 26,284
create_circular_range_50 circular 50 81,998 2,237 12,195
create_circular_range_100 circular 100 143,330 2,569 6,979
create_circular_range_255 circular 255 277,505 40,670 3,605

Execution

Test Name Type Range (Distance) Time (ns/iter) Variability (± ns) Iterations per Second (iters/s)
execute_linear_range_10 linear 10 95 1 10,526,316
execute_linear_range_100 linear 100 1,169 32 855,048
execute_linear_range_255 linear 255 2,783 349 359,323
execute_45_degrees_sector_range_10 radial_45 10 602 6 1,660,798
execute_45_degrees_sector_range_100 radial_45 100 23,884 666 41,866
execute_45_degrees_sector_range_255 radial_45 255 60,248 897 16,600
execute_90_degrees_sector_range_10 radial_90 10 1,227 29 815,073
execute_90_degrees_sector_range_100 radial_90 100 47,821 972 20,914
execute_90_degrees_sector_range_255 radial_90 255 121,197 25,467 8,250
execute_circular_range_3 circular 3 920 77 1,086,957
execute_circular_range_5 circular 5 2,034 123 491,699
execute_circular_range_10 circular 10 5,074 215 197,203
execute_circular_range_25 circular 25 27,759 923 36,020
execute_circular_range_50 circular 50 92,329 1,405 10,828
execute_circular_range_100 circular 100 199,025 3,846 5,025
execute_circular_range_255 circular 255 812,311 28,281 1,231

Navigable Collision

Test Name Type Range (Distance) Time (ns/iter) Variability (± ns) Iterations per Second (iters/s)
check_1million_obstacles_against_line_range_15 linear 15 44 1 22,727,273
check_1million_obstacles_against_line_range_50 linear 50 267 8 3,745,318
check_1million_obstacles_against_line_range_100 linear 100 585 15 1,709,402
check_1million_obstacles_against_line_range_255 linear 255 1,699 38 588,581
check_1million_obstacles_against_45_degrees_sector_range_15 radial_45 15 612 21 1,633,987
check_1million_obstacles_against_45_degrees_sector_range_50 radial_45 50 6,660 1,812 150,150
check_1million_obstacles_against_45_degrees_sector_range_100 radial_45 100 17,077 612 58,545
check_1million_obstacles_against_45_degrees_sector_range_255 radial_45 255 53,496 8,487 18,692
check_1million_obstacles_against_90_degrees_sector_range_15 radial_90 15 1,339 117 746,808
check_1million_obstacles_against_90_degrees_sector_range_50 radial_90 50 13,385 405 74,706
check_1million_obstacles_against_90_degrees_sector_range_100 radial_90 100 40,414 1,383 24,742
check_1million_obstacles_against_90_degrees_sector_range_255 radial_90 255 108,612 4,525 9,209
check_1million_obstacles_against_circle_range_15 circular 15 5,136 55 194,748
check_1million_obstacles_against_circle_range_50 circular 50 67,538 2,029 14,810
check_1million_obstacles_against_circle_range_100 circular 100 161,176 3,945 6,206
check_1million_obstacles_against_circle_range_255 circular 255 437,340 10,785 2,287

This README format clearly sections out the features, example usage, and benchmarks, providing a comprehensive guide for anyone looking to integrate the ryot_trajectories crate into their projects.

Dependencies

~19–61MB
~1M SLoC