Skip to main content

Rigid-bodies

The real-time simulation of rigid-bodies subjected to forces and contacts is the main feature of a physics engine for video-games, robotics, or animation. Rigid-bodies are typically used to simulate the dynamics of non-deformable solids as well as to integrate the trajectory of solids which velocities are controlled by the user (e.g. moving platforms). On the other hand, rigid-bodies are not enough to simulate, e.g., cars, ragdolls, or robotic systems, as those use-cases require adding restrictions on the relative motion between their parts using joints.

Note that rigid-bodies are only responsible for the dynamics and kinematics of the solid. Colliders can be attached to a rigid-body to specify its shape and enable collision-detection. A rigid-body without collider attached to it will not be affected by contacts (because there is no shape to compute contact against).

Creation and insertion​

A rigid-body is created from a R3RigidBodyDesc description. This plain structure must first be initialized by one of its constructors (which set the default value of every field), then any of its fields can be modified before it is inserted into the physics world with r3InsertRigidBody. The world copies the description and returns the R3RigidBodyHandle identifying the new rigid-body, which is then given to every function reading or modifying it.

info

The following example shows several fields that can be set to customize the rigid-body being built. The input values are just random so using this example as-is will not lead to a useful result.

// The world that will contain our rigid-bodies.
R2World *world = r2NewWorld();

// Description of a fixed rigid-body.
R2RigidBodyDesc fixed_desc = r2FixedRigidBodyDesc();
// Description of a dynamic rigid-body.
R2RigidBodyDesc dynamic_desc = r2DynamicRigidBodyDesc();
// Description of a kinematic rigid-body controlled at the velocity level.
R2RigidBodyDesc kinematic_velocity_desc = r2KinematicVelocityBasedRigidBodyDesc();
// Description of a kinematic rigid-body controlled at the position level.
R2RigidBodyDesc kinematic_position_desc = r2KinematicPositionBasedRigidBodyDesc();

R2RigidBodyDesc rigid_body = r2DynamicRigidBodyDesc();
// The body type: R2_DYNAMIC, R2_FIXED, R2_KINEMATIC_VELOCITY_BASED, or R2_KINEMATIC_POSITION_BASED.
// Default: the type of the constructor used to initialize the description.
rigid_body.bodyType = R2_DYNAMIC;
// The rigid body translation.
// Default: zero vector.
rigid_body.position.translation = r2Vector(0.0, 5.0);
// The rigid body rotation.
// Default: no rotation.
rigid_body.position.rotation = r2Rotation(5.0);
// The rigid body position. Will override the translation and rotation set above.
// Default: the identity pose.
rigid_body.position = r2Pose(r2Vector(1.0, 2.0), r2Rotation(0.4));
// The linear velocity of this body.
// Default: zero velocity.
rigid_body.linvel = r2Vector(1.0, 2.0);
// The angular velocity of this body.
// Default: zero velocity.
rigid_body.angvel = 2.0;
// The scaling factor applied to the gravity affecting the rigid-body.
// Default: 1.0
rigid_body.gravityScale = 0.5;
// Whether or not this body can sleep.
// Default: 1
rigid_body.canSleep = 1;
// Whether or not CCD is enabled for this rigid-body.
// Default: 0
rigid_body.ccdEnabled = 0;
// All done, actually create the rigid-body and insert it into the world.
R2RigidBodyHandle rigid_body_handle = r2InsertRigidBody(world, &rigid_body);

All the fields are optional. The only calls that are required are r3DynamicRigidBodyDesc(), r3FixedRigidBodyDesc(), r3KinematicVelocityBasedRigidBodyDesc(), or r3KinematicPositionBasedRigidBodyDesc(), to initialize the description, and r3InsertRigidBody to actually create the rigid-body. The rigid-body is removed from the world with r3RemoveRigidBody: its last argument indicates if its colliders must be removed as well (if it is 0, they are kept as colliders without parent).

info

Typically, the inertia and center of mass are automatically set to the inertia and center of mass resulting from the shapes of the colliders attached to the rigid-body. But they can also be set manually.

Rigid-body type​

There are four types of rigid-bodies, identified by the constants stored in the bodyType field of R3RigidBodyDesc:

  • R3_DYNAMIC: Indicates that the body is affected by external forces and contacts.
  • R3_FIXED: Indicates the body cannot move. It acts as if it has an infinite mass and will not be affected by any force. It will continue to collide with dynamic bodies but not with fixed nor with kinematic bodies. This is typically used for the ground or for temporarily freezing a body.
  • R3_KINEMATIC_POSITION_BASED: Indicates that the body position must not be altered by the physics engine. The user is free to set its next position and the body velocity will be deduced at each update accordingly to ensure a realistic behavior of dynamic bodies in contact with it. This is typically used for moving platforms, elevators, etc.
  • R3_KINEMATIC_VELOCITY_BASED: Indicates that the body velocity must not be altered by the physics engine. The user is free to set its velocity and the next body position will be deduced at each update accordingly to ensure a realistic behavior of dynamic bodies in contact with it. This is typically used for moving platforms, elevators, etc.

Both position-based and velocity-based kinematic bodies are mostly the same. Choosing between both is mostly a matter of preference between position-based control and velocity-based control.

Note that a fifth type exists, the soft frame, which is reserved to the rigid-bodies Rapier creates to give a frame to the soft-bodies: their pose is computed from the particles of the soft-body, and they aren't meant to be created by hand.

The type of a rigid-body can be read with r3RigidBody_BodyType (or tested with r3RigidBody_IsDynamic, r3RigidBody_IsFixed, and r3RigidBody_IsKinematic), and modified after its creation with r3RigidBody_SetBodyType.

info

The whole point of kinematic bodies is to let the user have total control over their trajectory. This means that kinematic bodies will simply ignore any contact force and go through walls and the ground. In other words: if you tell the kinematic to go somewhere, it will go there, no questions asked.

Taking obstacles into account needs to be done manually either by using scene queries to detect nearby obstacles, or by using the built-in character controller.

Position​

The position of a rigid-body represents its location (translation) in 2D or 3D world-space, as well as its orientation (rotation). Its translational part is represented as a vector (R3Vector) and its rotational part as a unit quaternion (R3Rotation) in 3D, or as an angle in radians (R2Rotation) in 2D. Both are combined into a pose (the R3Pose structure).

The position of a rigid-body can be set when creating it. It can also be set after its creation as illustrated below.

warning

Directly changing the position of a rigid-body is equivalent to teleporting it: this is a not a physically realistic action! Teleporting a dynamic or kinematic bodies may result in odd behaviors especially if it teleports into a space occupied by other objects. For dynamic bodies, forces, impulses, or velocity modification should be preferred. For kinematic bodies, see the discussion after the examples below.

/* Set the position when the rigid-body is created. */
R2RigidBodyDesc rigid_body = r2DynamicRigidBodyDesc();
// The rigid body translation.
// Default: zero vector.
rigid_body.position.translation = r2Vector(0.0, 5.0);
// The rigid body rotation.
// Default: no rotation.
rigid_body.position.rotation = r2Rotation(5.0);
// The rigid body position. Will override the translation and rotation set above.
// Default: the identity pose.
rigid_body.position = r2Pose(r2Vector(1.0, 2.0), r2Rotation(0.4));
/* Set the position after the rigid-body creation. */
// The last `1` argument makes sure the rigid-body is awake.
r2RigidBody_SetTranslation(rigid_body_handle, r2Vector(0.0, 5.0), 1);
r2RigidBody_SetRotation(rigid_body_handle, r2Rotation(0.2), 1);
R2Vector translation = r2RigidBody_Translation(rigid_body_handle);
R2Rotation rotation = r2RigidBody_Rotation(rigid_body_handle);
assert(translation.x == 0.0 && translation.y == 5.0);
assert(rotation.angle == (R2Real)0.2);

r2RigidBody_SetPosition(rigid_body_handle, r2Pose(r2Vector(1.0, 2.0), r2Rotation(0.4)), 1);
R2Pose position = r2RigidBody_Position(rigid_body_handle);
assert(position.translation.x == 1.0 && position.translation.y == 2.0);
assert(position.rotation.angle == (R2Real)0.4);

In order to move a dynamic rigid-body it is strongly discouraged to set its position directly as it may results in weird behaviors: it's as if the rigid-body teleports itself, which is a non-physical behavior. For dynamic bodies, it is recommended to either set its velocity or to apply forces or impulses.

For velocity-based kinematic bodies, it is recommended to set its velocity instead of setting its position directly. For position-based kinematic bodies, it is recommended to use the special functions:

  • r3RigidBody_SetNextKinematicRotation
  • r3RigidBody_SetNextKinematicTranslation
  • r3RigidBody_SetNextKinematicPosition (for both at once)

These functions will let the physics pipeline compute the fictitious velocity of the position-based kinematic body for more realistic interactions with other rigid-bodies. These functions won't immediately modify the position of the kinematic body itself. The position of the kinematic body will be automatically set to these values during the next physics pipeline update (the pending pose can be read with r3RigidBody_NextPosition).

Velocity​

The velocity of a dynamic rigid-body controls how fast it is moving in time. The velocity is applied at the center-of-mass of the rigid-body, and is composed of two independent parts:

  1. The linear velocity is specified as a vector representing the direction and magnitude of the movement.
  2. In 3D, the angular velocity is given as a vector representing the rotation axis multiplied by the rotation angular speed in rad/s (axis-angle representation). In 2D, the angular velocity is given as a real representing the angular speed in rad/s.
info

The velocity is only relevant to dynamic rigid-bodies. It has no effect on fixed rigid-bodies, and the velocity of kinematic rigid-bodies are automatically computed at each timestep based on their next kinematic positions.

The velocity of a rigid-body is automatically updated by the physics pipeline after taking forces, contacts, and joints into account. It can be set when the rigid-body is created or after its creation:

/* Set the velocities when the rigid-body is created. */
R2RigidBodyDesc rigid_body = r2DynamicRigidBodyDesc();
// The linear velocity of this body.
// Default: zero velocity.
rigid_body.linvel = r2Vector(1.0, 3.0);
// The angular velocity of this body.
// Default: zero velocity.
rigid_body.angvel = 3.0;
/* Set the velocities after the rigid-body creation. */
// The last `1` argument makes sure the rigid-body is awake.
r2RigidBody_SetLinvel(rigid_body_handle, r2Vector(1.0, 3.0), 1);
r2RigidBody_SetAngvel(rigid_body_handle, 3.0, 1);
R2Vector linvel = r2RigidBody_Linvel(rigid_body_handle);
assert(linvel.x == 1.0 && linvel.y == 3.0);
assert(r2RigidBody_Angvel(rigid_body_handle) == 3.0);

Alternatively, the velocity of a dynamic rigid-body can be altered indirectly by applying a force or an impulse.

Gravity​

Gravity is such a common force that it is implemented as a special case (even if it could easily be implemented by the user using force application). The gravity is set with r3SetGravity (and read with r3Gravity) and can be modified at will. Note however that a change of gravity won't automatically wake-up the sleeping bodies so keep in mind that you may want to wake them up manually before a gravity change.

note

Because fixed and kinematic bodies are immune to forces, they are not affected by gravity.

info

A rigid-body with no mass will not be affected by gravity either. So if your rigid-body doesn't fall when you expected it to, make sure it has a mass set explicitly, or has at least one collider with non-zero density attached to it.

It is possible to change the way gravity affects a specific rigid-body by setting the rigid-body's gravity scale to a value other than 1.0. The magnitude of the gravity applied to this body will be multiplied by this scaling factor. Therefore, a gravity scale set to 0.0 will disable gravity for the rigid-body whereas a gravity scale set to 2.0 will make it twice as strong. A negative value will flip the direction of the gravity for this rigid-body.

This gravity scale factor can be set when the rigid-body is created or after its creation:

/* Set the gravity scale when the rigid-body is created. */
R2RigidBodyDesc rigid_body = r2DynamicRigidBodyDesc();
// Divide by 2 the strength of gravity for this rigid-body.
rigid_body.gravityScale = 0.5;
/* Set the gravity scale after the rigid-body creation. */
// The last `1` argument makes sure the rigid-body is awake.
r2RigidBody_SetGravityScale(rigid_body_handle, 0.5, 1);
assert(r2RigidBody_GravityScale(rigid_body_handle) == 0.5);

Forces and impulses​

In addition to gravity, it is possible to add custom forces (or torques) or apply impulses (or torque impulses) to dynamic rigid-bodies in order to make them move in specific ways. Forces affect the rigid-body's acceleration whereas impulses affect the rigid-body's velocity. They are both based on the familiar equations:

  • Forces: the acceleration change is equal to the force divided by the mass: Δa=m−1f\Delta{}a = m^{-1}f
  • Impulses: the velocity change is equal to the impulse divided by the mass: Δv=m−1i\Delta{}v = m^{-1}i

Forces can be added, and impulses can be applied, to a rigid-body after it has been created. Added forces are persistent across simulation steps, and can be cleared manually.

// The last `1` argument makes sure the rigid-body is awake.
r2RigidBody_ResetForces(rigid_body_handle, 1);  // Reset the forces to zero.
r2RigidBody_ResetTorques(rigid_body_handle, 1); // Reset the torques to zero.
r2RigidBody_AddForce(rigid_body_handle, r2Vector(0.0, 1000.0), 1);
r2RigidBody_AddTorque(rigid_body_handle, 100.0, 1);
r2RigidBody_AddForceAtPoint(rigid_body_handle, r2Vector(0.0, 1000.0), r2Vector(1.0, 2.0), 1);

r2RigidBody_ApplyImpulse(rigid_body_handle, r2Vector(0.0, 1000.0), 1);
r2RigidBody_ApplyTorqueImpulse(rigid_body_handle, 100.0, 1);
r2RigidBody_ApplyImpulseAtPoint(rigid_body_handle, r2Vector(0.0, 1000.0), r2Vector(1.0, 2.0), 1);

The forces and torques added with r3RigidBody_AddForce, r3RigidBody_AddTorque, and r3RigidBody_AddForceAtPoint are accumulated until they are reset with r3RigidBody_ResetForces and r3RigidBody_ResetTorques. Their current sum can be read with r3RigidBody_UserForce and r3RigidBody_UserTorque. The impulses, on the other hand, modify the velocity of the rigid-body immediately. The points given to r3RigidBody_AddForceAtPoint and r3RigidBody_ApplyImpulseAtPoint are expressed in world-space.

info

Keep in mind that a dynamic rigid-body with a zero mass won't be affected by a linear force/impulse, and a rigid-body with a zero angular inertia won't be affected by torques/torque impulses. So if your force doesn't appear to do anything, make sure that:

  1. The rigid-body is dynamic.

  2. It is strong enough to make the rigid-body move (try a very large value and see if it does something).

  3. The rigid-body has a non-zero mass or angular inertia either because they were set explicitly, or because they were computed automatically from colliders with non-zero densities.

  4. The rigid-body is awake (by waking it up manually with r3RigidBody_WakeUp or setting the last wake_up argument to 1).

Mass properties​

The mass properties of a rigid-body is composed of three parts:

  • The mass which determines the resistance of the rigid-body wrt. linear movements. A high mass implies that larger forces are needed to make the rigid-body translate.
  • The angular inertia determines the resistance of the rigid-body wrt. the angular movements. A high angular inertia implies that larger torques are needed to make the rigid-body rotate.
  • The center-of-mass determines relative to what points torques are applied to the rigid-body.
note

Zero is a special value for masses and angular inertia. A mass equal to zero is interpreted as an infinite mass. An angular inertia equal to zero is interpreted as an infinite angular inertia. Therefore, a rigid-body with a mass equal to zero will not be affected by any force, and a rigid-body with an angular inertia equal to zero will not be affected by any torque.

Computing the mass and angular-inertia can often be difficult because they depend on the geometric shape of the object being simulated. This is why they are automatically computed by Rapier when a collider is attached to the rigid-body: the collider add its own mass and angular-inertia contribution (computed based on the collider's shape and density) to the rigid-body it is attached to:

R2RigidBodyDesc rigid_body = r2DynamicRigidBodyDesc();
R2RigidBodyHandle rigid_body_handle = r2InsertRigidBody(world, &rigid_body);
// The default density is 1.0, we are setting 2.0 for this example.
R2ColliderDesc collider = r2BallColliderDesc(1.0);
collider.density = 2.0;
// When the collider is attached, the rigid-body's mass and angular
// inertia is automatically updated to take the collider into account.
r2InsertCollider(rigid_body_handle, &collider);

Alternatively, it is possible to set the mass properties of a rigid-body when it is created. Keep in mind that this won't prevent the colliders' contributions to be added to these values. So make sure to set the attached colliders' densities to zero if you want your explicit values to be the final mass-properties values.

/* Set the mass-properties when the rigid-body is created. */
R2RigidBodyDesc rigid_body = r2DynamicRigidBodyDesc();
rigid_body.additionalMass = 0.5;
// Sets both the mass and angular inertia at once (this overrides `additionalMass`).
rigid_body.useAdditionalMassProperties = 1;
rigid_body.additionalMassProperties = (R2MassProperties){
    .local_com = r2Vector(0.0, 1.0),
    .mass = 0.5,
    .principal_inertia = 0.3,
};
/* Set the mass-properties after the rigid-body creation. */
R2MassProperties mass_properties = {
    .local_com = r2Vector(0.0, 1.0),
    .mass = 0.5,
    .principal_inertia = 0.3,
};
// The last `1` argument makes sure the rigid-body is awake.
r2RigidBody_SetAdditionalMassProperties(rigid_body_handle, mass_properties, 1);

The additionalMass field of R3RigidBodyDesc only adds a mass (the angular inertia being scaled accordingly, based on the shapes of the colliders), whereas the additionalMassProperties field (taken into account only if useAdditionalMassProperties is set to 1) specifies the full mass-properties (mass, center-of-mass, and principal angular inertia) of type R3MassProperties. After the creation of the rigid-body, they are modified with r3RigidBody_SetAdditionalMass and r3RigidBody_SetAdditionalMassProperties respectively.

The resulting mass-properties (including the colliders' contributions) can be read with r3RigidBody_Mass, r3RigidBody_LocalCenterOfMass (in the local-space of the rigid-body), and r3RigidBody_CenterOfMass (in world-space). They are updated automatically by the physics engine but, if you need them right after modifying the colliders or the additional mass-properties of a rigid-body (without waiting for the next timestep), they can be updated manually with r3RigidBody_RecomputeMassPropertiesFromColliders.

Locking translations/rotations​

It is sometimes useful to prevent a rigid-body from rotating or translating. One typical use-case for locking rotations is to prevent a player modeled as a dynamic rigid-body from tilting. These kind of degree-of-freedom restrictions could be achieved by joints, but locking translations/rotations of a single rigid-body wrt. the cartesian coordinate axes can be done in a much more efficient and numerically stable way. That's why rigid-bodies have dedicated flags for this.

The locked axes are given by a bitmask combining the R3_LOCK_TRANSLATION_X, R3_LOCK_TRANSLATION_Y, R3_LOCK_TRANSLATION_Z, R3_LOCK_ROTATION_X, R3_LOCK_ROTATION_Y, and R3_LOCK_ROTATION_Z flags (in 2D, only the translations along X and Y, and the rotation around Z, are relevant). It is set with the lockedAxes field of R3RigidBodyDesc when the rigid-body is created, or with r3RigidBody_SetLockedAxes afterwards (and read with r3RigidBody_LockedAxes). The r3RigidBody_SetTranslationsLocked and r3RigidBody_SetRotationsLocked functions lock (or unlock) all the translations or all the rotations at once.

/* Lock translations/rotations when the rigid-body is created. */
R2RigidBodyDesc rigid_body = r2DynamicRigidBodyDesc();
rigid_body.lockedAxes = R2_LOCK_TRANSLATION_X | R2_LOCK_TRANSLATION_Y // prevent translations along all axes.
                        | R2_LOCK_ROTATION_Z;                         // prevent rotations.
/* Lock translations/rotations after the rigid-body creation. */
// The last `1` argument makes sure the rigid-body is awake.
r2RigidBody_SetTranslationsLocked(rigid_body_handle, 1, 1);
r2RigidBody_SetRotationsLocked(rigid_body_handle, 1, 1);

Damping​

Damping lets you slow down a rigid-body automatically. This can be used to achieve a wide variety of effects like fake air friction. Each rigid-body is given a linear damping coefficient (affecting its linear velocity) and an angular damping coefficient (affecting its angular velocity). Larger values of the damping coefficients lead to a stronger slow-downs. Their default values are 0.0 (no damping at all).

This damping coefficients can be set when the rigid-body is created or after its creation:

/* Set the damping coefficients when the rigid-body is created. */
R2RigidBodyDesc rigid_body = r2DynamicRigidBodyDesc();
rigid_body.linearDamping = 0.5;
rigid_body.angularDamping = 1.0;
/* Set the damping coefficients after the rigid-body creation. */
r2RigidBody_SetLinearDamping(rigid_body_handle, 0.5);
r2RigidBody_SetAngularDamping(rigid_body_handle, 1.0);
assert(r2RigidBody_LinearDamping(rigid_body_handle) == 0.5);
assert(r2RigidBody_AngularDamping(rigid_body_handle) == 1.0);

Dominance​

Dominance is a non-realistic, but sometimes useful, feature. It can be used to make one rigid-body immune to forces originating from contacts with some other bodies. For example this can be used to model a player represented as a dynamic rigid-body that cannot be "pushed back" by any, or some, other dynamic rigid-bodies part of the environment.

Each rigid-body is part of a dominance group in [-127; 127] (the default group is 0). If the colliders from two rigid-bodies are in contact, the one with the highest dominance will act as if it has an infinite mass, making it immune to the contact forces the other body would apply on it. If both bodies are part of the same dominance group, then their contacts will work in the usual way (both are affected by opposite forces with the same magnitude).

For example, if a dynamic body A is in the dominance group 10, and a dynamic body B in the dominance group -20, then a contact between a collider attached to A and a collider attached B will result in A remaining immobile and B being pushed by A (independently from their mass).

info

A non-dynamic rigid-body is always considered as being part of a dominance group greater than any dynamic rigid-body. This means that dynamic/fixed and dynamic/kinematic contacts will continue to work normally, independently from the dominance group they were given by the user.

The dominance group can be set when the rigid-body is created or after its creation:

/* Set the dominance group when the rigid-body is created. */
R2RigidBodyDesc rigid_body = r2DynamicRigidBodyDesc();
rigid_body.dominanceGroup = 10;
/* Set the dominance group after the rigid-body creation. */
r2RigidBody_SetDominanceGroup(rigid_body_handle, 10);
assert(r2RigidBody_DominanceGroup(rigid_body_handle) == 10);

Continuous collision detection​

Continuous Collision Detection (CCD) is used to make sure that fast-moving objects don't miss any contacts (a problem usually called tunneling). This is done by looking for collisions along the shapes motion: a rigid-body that moved fast during the timestep casts its colliders from their previous position to their new one, and its position is clamped to the first impact found this way. Its velocities are left untouched and the solver is responsible for preventing penetrations. Since the trajectory is clamped to some intermediate location along its path this technique is commonly called motion clamping.

Rapier applies CCD in two ways:

  • Every fast-moving dynamic rigid-body is swept against the fixed colliders and the soft-bodies of the scene. This is automatic and doesn't need to be enabled: this prevents a falling crate from going through the floor, or a projectile from going through a wall, etc. Note however that this is not enabled fol dynamic rigid-bodies with mesh-like shapes (triangle meshes, polylines, heightfields) as that would be too computationally expensive.
  • A dynamic rigid-body with CCD enabled (aka. a bullet) is swept against the kinematic and dynamic bodies as well. This is more expensive, therefore it is disabled by default and should be reserved to the objects that must not tunnel through moving obstacles. Note that two CCD-enabled objects might still tunel since the CCD resolution does currently not take both continuous motions into account simultaneously.
info

CCD takes action only if the rigid-body is moving fast relative to another collider. Therefore it is useless to enable it on fixed rigid-bodies and on rigid-bodies that are expected to move slowly.

The CCD feature, including the automatic sweeping of the fast dynamic bodies, can be fully disabled by setting the maximum number of CCD substeps to zero in the IntegrationParameters (the R3IntegrationParameters.maxCcdSubsteps field, whose default is 1). Larger values let a body resolve several successive impacts within a single timestep, at the cost of additional sweeps. It can also be modified directly with r3SetMaxCcdSubsteps.

Per-object CCD can be enabled when creating a rigid-body or after its creation:

/* Enable CCD when the rigid-body is created. */
R2RigidBodyDesc rigid_body = r2DynamicRigidBodyDesc();
rigid_body.ccdEnabled = 1;
/* Enable CCD after the rigid-body creation. */
r2RigidBody_SetCcdEnabled(rigid_body_handle, 1);
assert(r2RigidBody_IsCcdEnabled(rigid_body_handle));

Keep in mind that r3RigidBody_IsCcdEnabled only tells if CCD was enabled for this rigid-body. Whether a rigid-body is currently moving fast enough for CCD to take action (which is also the case of fast dynamic bodies without CCD enabled, since they are swept against the fixed colliders) is given by r3RigidBody_IsCcdActive.

Sleeping​

When a dynamic rigid-body doesn't move (or moves very slowly) during a few seconds, it will be marked as sleeping by the physics pipeline. Rigid-bodies marked as sleeping are no longer simulated by the physics engine until they are woken up. That way the physics engine doesn't waste any computational resources simulating objects that don't actually move. They are woken up automatically whenever another non-sleeping rigid-body starts interacting with them (either with a joint, or with one of its attached colliders generating contacts).

However, a sleeping rigid-body won't respond to any user action. This is why it is possible to wake-up the rigid-body manually with r3RigidBody_WakeUp (if its strong argument is 1, the rigid-body is guaranteed to stay awake for several timesteps, otherwise it may fall asleep again immediately). Some functions take an additional wake_up argument that, if set to 1, ensures that the rigid-body wakes up before the action takes place. For example:

  • r3RigidBody_AddForce(handle, force, 1) will wake-up the rigid-body before adding the force.
  • r3RemoveImpulseJoint(joint, 1) (resp. r3RemoveMultibodyJoint) will wake-up the two rigid-bodies attached by the removed joint.
  • r3RemoveCollider(collider, 1) will wake-up the rigid-body the removed collider is attached to.

Unless you want to achieve special effects, it is recommended to always set the wake_up argument to 1. One example of case where setting the argument of wake_up to 0 makes sense is to simulate a custom constant gravity with r3RigidBody_AddForce(handle, force, 0). This will result in the force being added to the rigid-body, but will allow the rigid-body to fall asleep if it reaches a dynamic equilibrium.

Whether a rigid-body is sleeping is given by r3RigidBody_IsSleeping, and it can be put to sleep manually with r3RigidBody_Sleep. A rigid-body can be prevented from ever sleeping by setting the canSleep field of its R3RigidBodyDesc to 0, or created already asleep by setting its sleeping field to 1.

Solver settings​

The accuracy of the constraints solver is configured for the whole world by the integration parameters. However, in some cases, part of the simulation might need more fine-grained control. For example, an articulated robot, or a stack involving large mass ratios, might require more solver iterations. This is why a rigid-body can ask for additional solver iterations (either substeps, or internal steps) resulting in the island it belongs to (the bodies it is connected to by contacts and joints) to run with a higher accuracy without hurting the performances of other islands.

Two other settings affect how a rigid-body is integrated. The soft-CCD prediction distance makes the body generate predictive contacts ahead of its own path, which is a cheaper alternative to CCD for the objects that are thin or moderately fast (large values impact the performances badly by increasing significantly the number of collision pairs). Finally, the fast-rotation flag lets the body exceed the angular speed cap, which is enabled by default to keep CCD reliable.

These settings are given by the additionalSolverIterations, additionalPgsIterations, softCcdPrediction, and allowFastRotation fields of R3RigidBodyDesc. After the creation of the rigid-body, they can be modified with r3RigidBody_SetAdditionalSolverIterations, r3RigidBody_SetAdditionalPgsIterations, r3RigidBody_SetSoftCcdPrediction, and r3RigidBody_SetAllowFastRotation (and read with r3RigidBody_AdditionalSolverIterations, r3RigidBody_AdditionalPgsIterations, r3RigidBody_SoftCcdPrediction, and r3RigidBody_IsFastRotationAllowed).

/* Give a rigid-body more solver accuracy than the rest of the scene. */
R2RigidBodyDesc rigid_body = r2DynamicRigidBodyDesc();
// Extra substeps run for the whole island component this body belongs to.
rigid_body.additionalSolverIterations = 4;
// Extra internal PGS iterations run per substep for that same component.
rigid_body.additionalPgsIterations = 2;
// Predictive contacts generated up to that distance ahead of the body's path: a cheaper
// alternative to CCD for slow-but-thin or moderately fast objects.
rigid_body.softCcdPrediction = 0.5;
// Let the body exceed the angular speed cap, e.g. for a wheel.
rigid_body.allowFastRotation = 1;
note

In 3D, the gyroscopic forces of a rigid-body can be disabled as well by setting the gyroscopicForcesEnabled field of its R3RigidBodyDesc to 0, or with r3RigidBody_SetGyroscopicForcesEnabled after its creation. When enabled (the default), they give the more realistic behaviors of a spinning solid, e.g., the precession of a spinning top or the Dzhanibekov effect. Disabling them is only recommended if they represent a measurable overhead in your simulation.

User-data​

Each rigid-body can be given a user-defined data of type R3UserData: a 128-bits integer split into its low and high 64-bits halves. This integer can have any value and is never used/modified by the physics-engine. This can for example be useful to store an index (or a pointer converted to an integer) referencing your own data associated to the rigid-body, or to add some custom data for custom contact filtering/modification.

This user-data can be set when the rigid-body is created or after its creation:

/* Set the user-data when the rigid-body is created. */
R2RigidBodyDesc rigid_body = r2DynamicRigidBodyDesc();
// The 128 bits of the user-data are split into its `low` and `high` 64 bits.
rigid_body.userData.low = 42;
/* Set the user-data after the rigid-body creation. */
R2UserData user_data = {.low = 42, .high = 0};
r2RigidBody_SetUserData(rigid_body_handle, user_data);
assert(r2RigidBody_UserData(rigid_body_handle).low == 42);