This library contains the implementations of the GJK and EPA algorithms. They are used a lot in the domain of computer graphics, to:
- detect collisions between pairs of convex shapes
- get the closest pair of points between pairs of convex shapes
- get the penetration vector between pairs of convex shapes
With respect to similar implementations, this one allows to easily:
- account for roto-traslation, without explicitly computes the vertices coordinates after the roto-traslation
- use a template container to describe your the convex shape, using the 3D coordinate representation of your favourite algebra library (Eigen, etc...)
Before diving into the code, it is strongly recomended to have a look at the documentation.
This library is stand-alone and completely cross platform. Use CMake to configure the project.
Haven't yet left a star? Do it now! :). Performing proximity queries with this package is pretty easy. Let's assume to have defined a pair of convex shapes:
// build the first convex shape ... have a look at the samples to understand
// how to do it
const flx::shape::ConvexShape &shapeA = ;
// build the second convex shape ... have a look at the samples to understand
// how to do it
const flx::shape::ConvexShape &shapeB = ;We can simply check the presence of a collision:
bool result = flx::is_collision_present(shapeA, shapeB);
// this will call only the initial iterations of the GJK (see the
// documentation) required to answer the query.If we suspect the shapes ARE in collision, we can get the penetration vector (see the documentation) by calling:
std::optional<flx::CoordinatePair> result =
flx::get_penetration_info(shapeA, shapeB);
// this will call the initial iterations of the GJK + the EPA ones in
// case the shapes are in collision.
//
// On the contrary, when the shapes are not in collision a std::nullopt is
// actually returned, and only the initial iterations of the GJK are done.
//
// If the result is not null, you can access the extremals of the
// penetration vector:
if (std::nullopt != result) {
const hull::Coordinate &point_on_shapeA = result->point_in_shape_a;
const hull::Coordinate &point_on_shapeB = result->point_in_shape_b;
}If we suspect the shapes ARE NOT in collision, we can get the closest pair of points on the 2 shapes by calling:
std::optional<flx::CoordinatePair> result =
flx::get_closest_points(shapeA, shapeB);
// this will call only initial iterations of the GJK + refining iterations
// of the GJK, evolving the plex till finding the closest region of the
// Minkowski difference to the origin. This is actually done only in case
// the shapes are not in collision.
//
// On the contrary, when the shapes are in collision a std::nullopt is
// actually returned, and only the initial iterations of the GJK are done.
//
// If the result is not null, you can access the the closest pair from the
// result:
if (std::nullopt != result) {
const hull::Coordinate &closest_point_on_shapeA =
result->point_in_shape_a;
const hull::Coordinate &closest_point_on_shapeB =
result->point_in_shape_b;
}...or, we can ask to perform a generic complex query that leads to call the EPA or the finishing iterations of the GJK to respectively compute the penetration vector or the closest pair:
flx::QueryResult result =
flx::get_closest_points_or_penetration_info(shapeA, shapeB);
// and then access the results (which might have 2 different meanings)
bool result_meaning = result.is_closest_pair_or_penetration_info;
if (result_meaning) {
// is the closest pair
const hull::Coordinate &closest_point_on_shapeA =
result.result.point_in_shape_a;
const hull::Coordinate &closest_point_on_shapeB =
result.result.point_in_shape_b;
} else {
// extremals of the penetration vector
const hull::Coordinate &point_on_shapeA = result.result.point_in_shape_a;
const hull::Coordinate &point_on_shapeB = result.result.point_in_shape_b;
}The relevant code is contained in the src folder, while samples contains examples showing how to use this library. In particular, after running each sample, some .json files will be produced storing the results. A python script can be later used to visualize the results. The syntax to use will appear in the console after computing each individual result.
Haven't yet left a star? Do it now! :).
To consume this library you can rely on CMake. More precisely, You can fetch this package and link to the GJK-EPA library:
include(FetchContent)
FetchContent_Declare(
gjk_epa
GIT_REPOSITORY https://github.com/andreacasalino/Flexible-GJK-and-EPA
GIT_TAG master
)
FetchContent_MakeAvailable(gjk_epa)and then link to the GJK-EPA library:
target_link_libraries(${TARGET_NAME}
GJK-EPA
)