use crate::dynamics::solver::DeltaVel; use crate::dynamics::{ GenericJoint, IntegrationParameters, JointGraphEdge, JointIndex, JointParams, RigidBody, }; use crate::math::{AngularInertia, Dim, Isometry, Real, SpacialVector, Vector, DIM}; use crate::parry::math::{AngDim, SpatialVector}; use crate::utils::{WAngularInertia, WCross, WCrossMatrix}; #[cfg(feature = "dim2")] use na::{Matrix3, Vector3}; #[cfg(feature = "dim3")] use na::{Matrix6, Vector6, U3}; #[derive(Debug)] pub(crate) struct GenericVelocityConstraint { mj_lambda1: usize, mj_lambda2: usize, joint_id: JointIndex, impulse: SpacialVector, pos_impulse: SpacialVector, max_positive_impulse: SpatialVector, max_negative_impulse: SpatialVector, #[cfg(feature = "dim3")] inv_lhs: Matrix6, // FIXME: replace by Cholesky. #[cfg(feature = "dim3")] rhs: Vector6, #[cfg(feature = "dim2")] inv_lhs: Matrix3, // FIXME: replace by Cholesky. #[cfg(feature = "dim2")] rhs: Vector3, pos_rhs: Vector6, im1: Real, im2: Real, ii1: AngularInertia, ii2: AngularInertia, ii1_sqrt: AngularInertia, ii2_sqrt: AngularInertia, r1: Vector, r2: Vector, } impl GenericVelocityConstraint { #[inline(always)] pub fn compute_mass_matrix( joint: &GenericJoint, im1: Real, im2: Real, ii1: AngularInertia, ii2: AngularInertia, r1: Vector, r2: Vector, velocity_solver: bool, ) -> Matrix6 { let rmat1 = r1.gcross_matrix(); let rmat2 = r2.gcross_matrix(); #[allow(unused_mut)] // For 2D let mut lhs; #[cfg(feature = "dim3")] { let lhs00 = ii1.quadform(&rmat1).add_diagonal(im1) + ii2.quadform(&rmat2).add_diagonal(im2); let lhs10 = ii1.transform_matrix(&rmat1) + ii2.transform_matrix(&rmat2); let lhs11 = (ii1 + ii2).into_matrix(); // Note that Cholesky only reads the lower-triangular part of the matrix // so we don't need to fill lhs01. lhs = Matrix6::zeros(); lhs.fixed_slice_mut::(0, 0) .copy_from(&lhs00.into_matrix()); lhs.fixed_slice_mut::(3, 0).copy_from(&lhs10); lhs.fixed_slice_mut::(3, 3).copy_from(&lhs11); // Adjust the mass matrix to take force limits into account. // If a DoF has a force limit, then we need to make its // constraint independent from the others because otherwise // the force clamping will cause errors to propagate in the // other constraints. if velocity_solver { for i in 0..6 { if joint.max_negative_impulse[i] > -Real::MAX || joint.max_positive_impulse[i] < Real::MAX { let diag = lhs[(i, i)]; lhs.column_mut(i).fill(0.0); lhs.row_mut(i).fill(0.0); lhs[(i, i)] = diag; } } } } // In 2D we just unroll the computation because // it's just easier that way. #[cfg(feature = "dim2")] { let m11 = im1 + im2 + rmat1.x * rmat1.x * ii1 + rmat2.x * rmat2.x * ii2; let m12 = rmat1.x * rmat1.y * ii1 + rmat2.x * rmat2.y * ii2; let m22 = im1 + im2 + rmat1.y * rmat1.y * ii1 + rmat2.y * rmat2.y * ii2; let m13 = rmat1.x * ii1 + rmat2.x * ii2; let m23 = rmat1.y * ii1 + rmat2.y * ii2; let m33 = ii1 + ii2; lhs = Matrix3::new(m11, m12, m13, m12, m22, m23, m13, m23, m33) } // NOTE: we don't use Cholesky in 2D because we only have a 3x3 matrix // for which a textbook inverse is still efficient. #[cfg(feature = "dim2")] return lhs.try_inverse().expect("Singular system."); #[cfg(feature = "dim3")] return lhs.cholesky().expect("Singular system.").inverse(); } pub fn from_params( params: &IntegrationParameters, joint_id: JointIndex, rb1: &RigidBody, rb2: &RigidBody, joint: &GenericJoint, ) -> Self { let anchor1 = rb1.position * joint.local_anchor1; let anchor2 = rb2.position * joint.local_anchor2; let im1 = rb1.effective_inv_mass; let im2 = rb2.effective_inv_mass; let ii1 = rb1.effective_world_inv_inertia_sqrt.squared(); let ii2 = rb2.effective_world_inv_inertia_sqrt.squared(); let r1 = anchor1.translation.vector - rb1.world_com.coords; let r2 = anchor2.translation.vector - rb2.world_com.coords; let lin_dvel = -rb1.linvel - rb1.angvel.gcross(r1) + rb2.linvel + rb2.angvel.gcross(r2); let ang_dvel = -rb1.angvel + rb2.angvel; let inv_lhs = Self::compute_mass_matrix(joint, im1, im2, ii1, ii2, r1, r2, true); #[cfg(feature = "dim2")] let dvel = Vector3::new(lin_dvel.x, lin_dvel.y, ang_dvel); #[cfg(feature = "dim3")] let dvel = Vector6::new( lin_dvel.x, lin_dvel.y, lin_dvel.z, ang_dvel.x, ang_dvel.y, ang_dvel.z, ); let target_linvel = anchor2 * joint.target_velocity.xyz(); let target_angvel = anchor2 * joint.target_velocity.fixed_rows::(DIM).into_owned(); let target_vel = Vector6::new( target_linvel.x, target_linvel.y, target_linvel.z, target_angvel.x, target_angvel.y, target_angvel.z, ); let rhs = dvel - dvel.sup(&target_vel).inf(&target_vel); let delta_pos = Isometry::from_parts( anchor2.translation * anchor1.translation.inverse(), anchor2.rotation * anchor1.rotation.inverse(), ); let lin_dpos = delta_pos.translation.vector; let ang_dpos = delta_pos.rotation.scaled_axis(); let dpos = Vector6::new( lin_dpos.x, lin_dpos.y, lin_dpos.z, ang_dpos.x, ang_dpos.y, ang_dpos.z, ); let err = dpos - dpos.sup(&joint.min_position).inf(&joint.max_position); let pos_rhs = err * params.inv_dt() * params.joint_erp; let impulse = (joint.impulse * params.warmstart_coeff) .inf(&joint.max_positive_impulse) .sup(&joint.max_negative_impulse); GenericVelocityConstraint { joint_id, mj_lambda1: rb1.active_set_offset, mj_lambda2: rb2.active_set_offset, im1, im2, ii1, ii2, ii1_sqrt: rb1.effective_world_inv_inertia_sqrt, ii2_sqrt: rb2.effective_world_inv_inertia_sqrt, impulse, pos_impulse: na::zero(), max_positive_impulse: joint.max_positive_impulse, max_negative_impulse: joint.max_negative_impulse, inv_lhs, r1, r2, rhs, pos_rhs, } } pub fn warmstart(&self, mj_lambdas: &mut [DeltaVel]) { let mut mj_lambda1 = mj_lambdas[self.mj_lambda1 as usize]; let mut mj_lambda2 = mj_lambdas[self.mj_lambda2 as usize]; let lin_impulse = self.impulse.fixed_rows::(0).into_owned(); #[cfg(feature = "dim2")] let ang_impulse = self.impulse[2]; #[cfg(feature = "dim3")] let ang_impulse = self.impulse.fixed_rows::(3).into_owned(); mj_lambda1.linear += self.im1 * lin_impulse; mj_lambda1.angular += self .ii1_sqrt .transform_vector(ang_impulse + self.r1.gcross(lin_impulse)); mj_lambda2.linear -= self.im2 * lin_impulse; mj_lambda2.angular -= self .ii2_sqrt .transform_vector(ang_impulse + self.r2.gcross(lin_impulse)); mj_lambdas[self.mj_lambda1 as usize] = mj_lambda1; mj_lambdas[self.mj_lambda2 as usize] = mj_lambda2; } pub fn solve(&mut self, mj_lambdas: &mut [DeltaVel]) { return; let mut mj_lambda1 = mj_lambdas[self.mj_lambda1 as usize]; let mut mj_lambda2 = mj_lambdas[self.mj_lambda2 as usize]; let ang_vel1 = self.ii1_sqrt.transform_vector(mj_lambda1.angular); let ang_vel2 = self.ii2_sqrt.transform_vector(mj_lambda2.angular); let dlinvel = -mj_lambda1.linear - ang_vel1.gcross(self.r1) + mj_lambda2.linear + ang_vel2.gcross(self.r2); let dangvel = -ang_vel1 + ang_vel2; #[cfg(feature = "dim2")] let rhs = Vector3::new(dlinvel.x, dlinvel.y, dangvel) + self.rhs; #[cfg(feature = "dim3")] let dvel = Vector6::new( dlinvel.x, dlinvel.y, dlinvel.z, dangvel.x, dangvel.y, dangvel.z, ) + self.rhs; let new_impulse = (self.impulse + self.inv_lhs * dvel) .sup(&self.max_negative_impulse) .inf(&self.max_positive_impulse); let effective_impulse = new_impulse - self.impulse; self.impulse = new_impulse; let lin_impulse = effective_impulse.fixed_rows::(0).into_owned(); #[cfg(feature = "dim2")] let ang_impulse = effective_impulse[2]; #[cfg(feature = "dim3")] let ang_impulse = effective_impulse.fixed_rows::(3).into_owned(); mj_lambda1.linear += self.im1 * lin_impulse; mj_lambda1.angular += self .ii1_sqrt .transform_vector(ang_impulse + self.r1.gcross(lin_impulse)); mj_lambda2.linear -= self.im2 * lin_impulse; mj_lambda2.angular -= self .ii2_sqrt .transform_vector(ang_impulse + self.r2.gcross(lin_impulse)); mj_lambdas[self.mj_lambda1 as usize] = mj_lambda1; mj_lambdas[self.mj_lambda2 as usize] = mj_lambda2; } pub fn solve2( &mut self, mj_lambdas: &mut [DeltaVel], mj_lambdas_pos: &mut [DeltaVel], ) { let mut mj_lambda1 = mj_lambdas[self.mj_lambda1 as usize]; let mut mj_lambda2 = mj_lambdas[self.mj_lambda2 as usize]; let mut mj_lambda_pos1 = mj_lambdas_pos[self.mj_lambda1 as usize]; let mut mj_lambda_pos2 = mj_lambdas_pos[self.mj_lambda2 as usize]; /* * Solve velocity. */ let ang_vel1 = self.ii1_sqrt.transform_vector(mj_lambda1.angular); let ang_vel2 = self.ii2_sqrt.transform_vector(mj_lambda2.angular); let dlinvel = -mj_lambda1.linear - ang_vel1.gcross(self.r1) + mj_lambda2.linear + ang_vel2.gcross(self.r2); let dangvel = -ang_vel1 + ang_vel2; #[cfg(feature = "dim2")] let rhs = Vector3::new(dlinvel.x, dlinvel.y, dangvel) + self.rhs; #[cfg(feature = "dim3")] let dvel = Vector6::new( dlinvel.x, dlinvel.y, dlinvel.z, dangvel.x, dangvel.y, dangvel.z, ) + self.rhs; let new_impulse = (self.impulse + self.inv_lhs * dvel) .sup(&self.max_negative_impulse) .inf(&self.max_positive_impulse); let effective_impulse = new_impulse - self.impulse; self.impulse = new_impulse; let lin_impulse = effective_impulse.fixed_rows::(0).into_owned(); #[cfg(feature = "dim2")] let ang_impulse = effective_impulse[2]; #[cfg(feature = "dim3")] let ang_impulse = effective_impulse.fixed_rows::(3).into_owned(); mj_lambda1.linear += self.im1 * lin_impulse; mj_lambda1.angular += self .ii1_sqrt .transform_vector(ang_impulse + self.r1.gcross(lin_impulse)); mj_lambda2.linear -= self.im2 * lin_impulse; mj_lambda2.angular -= self .ii2_sqrt .transform_vector(ang_impulse + self.r2.gcross(lin_impulse)); /* * Solve positions. */ let ang_pos1 = self.ii1_sqrt.transform_vector(mj_lambda_pos1.angular); let ang_pos2 = self.ii2_sqrt.transform_vector(mj_lambda_pos2.angular); let dlinpos = -mj_lambda_pos1.linear - ang_pos1.gcross(self.r1) + mj_lambda_pos2.linear + ang_pos2.gcross(self.r2); let dangpos = -ang_pos1 + ang_pos2; #[cfg(feature = "dim3")] let dpos = Vector6::new( dlinpos.x, dlinpos.y, dlinpos.z, dangpos.x, dangpos.y, dangpos.z, ) + self.pos_rhs; let new_impulse = self.pos_impulse + self.inv_lhs * dpos; let effective_impulse = new_impulse - self.pos_impulse; self.pos_impulse = new_impulse; let lin_impulse = effective_impulse.fixed_rows::(0).into_owned(); #[cfg(feature = "dim2")] let ang_impulse = effective_impulse[2]; #[cfg(feature = "dim3")] let ang_impulse = effective_impulse.fixed_rows::(3).into_owned(); mj_lambda_pos1.linear += self.im1 * lin_impulse; mj_lambda_pos1.angular += self .ii1_sqrt .transform_vector(ang_impulse + self.r1.gcross(lin_impulse)); mj_lambda_pos2.linear -= self.im2 * lin_impulse; mj_lambda_pos2.angular -= self .ii2_sqrt .transform_vector(ang_impulse + self.r2.gcross(lin_impulse)); mj_lambdas[self.mj_lambda1 as usize] = mj_lambda1; mj_lambdas[self.mj_lambda2 as usize] = mj_lambda2; mj_lambdas_pos[self.mj_lambda1 as usize] = mj_lambda_pos1; mj_lambdas_pos[self.mj_lambda2 as usize] = mj_lambda_pos2; } pub fn writeback_impulses(&self, joints_all: &mut [JointGraphEdge]) { let joint = &mut joints_all[self.joint_id].weight; if let JointParams::GenericJoint(fixed) = &mut joint.params { fixed.impulse = self.impulse; } } } #[derive(Debug)] pub(crate) struct GenericVelocityGroundConstraint { mj_lambda2: usize, joint_id: JointIndex, impulse: SpacialVector, pos_impulse: SpacialVector, max_positive_impulse: SpatialVector, max_negative_impulse: SpatialVector, #[cfg(feature = "dim3")] inv_lhs: Matrix6, // FIXME: replace by Cholesky. #[cfg(feature = "dim3")] rhs: Vector6, #[cfg(feature = "dim2")] inv_lhs: Matrix3, // FIXME: replace by Cholesky. #[cfg(feature = "dim2")] rhs: Vector3, pos_rhs: Vector6, im2: Real, ii2: AngularInertia, ii2_sqrt: AngularInertia, r2: Vector, } impl GenericVelocityGroundConstraint { #[inline(always)] pub fn compute_mass_matrix( joint: &GenericJoint, im2: Real, ii2: AngularInertia, r2: Vector, velocity_solver: bool, ) -> Matrix6 { let rmat2 = r2.gcross_matrix(); #[allow(unused_mut)] // For 2D. let mut lhs; #[cfg(feature = "dim3")] { let lhs00 = ii2.quadform(&rmat2).add_diagonal(im2); let lhs10 = ii2.transform_matrix(&rmat2); let lhs11 = ii2.into_matrix(); // Note that Cholesky only reads the lower-triangular part of the matrix // so we don't need to fill lhs01. lhs = Matrix6::zeros(); lhs.fixed_slice_mut::(0, 0) .copy_from(&lhs00.into_matrix()); lhs.fixed_slice_mut::(3, 0).copy_from(&lhs10); lhs.fixed_slice_mut::(3, 3).copy_from(&lhs11); // Adjust the mass matrix to take force limits into account. // If a DoF has a force limit, then we need to make its // constraint independent from the others because otherwise // the force clamping will cause errors to propagate in the // other constraints. if velocity_solver { for i in 0..6 { if joint.max_negative_impulse[i] > -Real::MAX || joint.max_positive_impulse[i] < Real::MAX { let diag = lhs[(i, i)]; lhs.column_mut(i).fill(0.0); lhs.row_mut(i).fill(0.0); lhs[(i, i)] = diag; } } } } // In 2D we just unroll the computation because // it's just easier that way. #[cfg(feature = "dim2")] { let m11 = im2 + rmat2.x * rmat2.x * ii2; let m12 = rmat2.x * rmat2.y * ii2; let m22 = im2 + rmat2.y * rmat2.y * ii2; let m13 = rmat2.x * ii2; let m23 = rmat2.y * ii2; let m33 = ii2; lhs = Matrix3::new(m11, m12, m13, m12, m22, m23, m13, m23, m33) } #[cfg(feature = "dim2")] return lhs.try_inverse().expect("Singular system."); #[cfg(feature = "dim3")] return lhs.cholesky().expect("Singular system.").inverse(); } pub fn from_params( params: &IntegrationParameters, joint_id: JointIndex, rb1: &RigidBody, rb2: &RigidBody, joint: &GenericJoint, flipped: bool, ) -> Self { let (anchor1, anchor2) = if flipped { ( rb1.position * joint.local_anchor2, rb2.position * joint.local_anchor1, ) } else { ( rb1.position * joint.local_anchor1, rb2.position * joint.local_anchor2, ) }; let r1 = anchor1.translation.vector - rb1.world_com.coords; let im2 = rb2.effective_inv_mass; let ii2 = rb2.effective_world_inv_inertia_sqrt.squared(); let r2 = anchor2.translation.vector - rb2.world_com.coords; let inv_lhs = Self::compute_mass_matrix(joint, im2, ii2, r2, true); let lin_dvel = rb2.linvel + rb2.angvel.gcross(r2) - rb1.linvel - rb1.angvel.gcross(r1); let ang_dvel = rb2.angvel - rb1.angvel; #[cfg(feature = "dim2")] let dvel = Vector3::new(lin_dvel.x, lin_dvel.y, ang_dvel); #[cfg(feature = "dim3")] let dvel = Vector6::new( lin_dvel.x, lin_dvel.y, lin_dvel.z, ang_dvel.x, ang_dvel.y, ang_dvel.z, ); let target_linvel = anchor2 * joint.target_velocity.xyz(); let target_angvel = anchor2 * joint.target_velocity.fixed_rows::(DIM).into_owned(); let target_vel = Vector6::new( target_linvel.x, target_linvel.y, target_linvel.z, target_angvel.x, target_angvel.y, target_angvel.z, ); let mut rhs = dvel - dvel.sup(&target_vel).inf(&target_vel); let delta_pos = Isometry::from_parts( anchor2.translation * anchor1.translation.inverse(), anchor2.rotation * anchor1.rotation.inverse(), ); let lin_dpos = delta_pos.translation.vector; let ang_dpos = delta_pos.rotation.scaled_axis(); let dpos = Vector6::new( lin_dpos.x, lin_dpos.y, lin_dpos.z, ang_dpos.x, ang_dpos.y, ang_dpos.z, ); let err = dpos - dpos.sup(&joint.min_position).inf(&joint.max_position); let pos_rhs = err * params.inv_dt() * params.joint_erp; let impulse = (joint.impulse * params.warmstart_coeff) .inf(&joint.max_positive_impulse) .sup(&joint.max_negative_impulse); GenericVelocityGroundConstraint { joint_id, mj_lambda2: rb2.active_set_offset, im2, ii2, ii2_sqrt: rb2.effective_world_inv_inertia_sqrt, impulse, pos_impulse: na::zero(), max_positive_impulse: joint.max_positive_impulse, max_negative_impulse: joint.max_negative_impulse, inv_lhs, r2, rhs, pos_rhs, } } pub fn warmstart(&self, mj_lambdas: &mut [DeltaVel]) { let mut mj_lambda2 = mj_lambdas[self.mj_lambda2 as usize]; let lin_impulse = self.impulse.fixed_rows::(0).into_owned(); #[cfg(feature = "dim2")] let ang_impulse = self.impulse[2]; #[cfg(feature = "dim3")] let ang_impulse = self.impulse.fixed_rows::(3).into_owned(); mj_lambda2.linear -= self.im2 * lin_impulse; mj_lambda2.angular -= self .ii2_sqrt .transform_vector(ang_impulse + self.r2.gcross(lin_impulse)); mj_lambdas[self.mj_lambda2 as usize] = mj_lambda2; } pub fn solve(&mut self, mj_lambdas: &mut [DeltaVel]) { return; let mut mj_lambda2 = mj_lambdas[self.mj_lambda2 as usize]; let ang_vel2 = self.ii2_sqrt.transform_vector(mj_lambda2.angular); let dlinvel = mj_lambda2.linear + ang_vel2.gcross(self.r2); let dangvel = ang_vel2; #[cfg(feature = "dim2")] let rhs = Vector3::new(dlinvel.x, dlinvel.y, dangvel) + self.rhs; #[cfg(feature = "dim3")] let dvel = Vector6::new( dlinvel.x, dlinvel.y, dlinvel.z, dangvel.x, dangvel.y, dangvel.z, ) + self.rhs; let new_impulse = (self.impulse + self.inv_lhs * dvel) .sup(&self.max_negative_impulse) .inf(&self.max_positive_impulse); let effective_impulse = new_impulse - self.impulse; self.impulse = new_impulse; let lin_impulse = effective_impulse.fixed_rows::(0).into_owned(); #[cfg(feature = "dim2")] let ang_impulse = effective_impulse[2]; #[cfg(feature = "dim3")] let ang_impulse = effective_impulse.fixed_rows::(3).into_owned(); mj_lambda2.linear -= self.im2 * lin_impulse; mj_lambda2.angular -= self .ii2_sqrt .transform_vector(ang_impulse + self.r2.gcross(lin_impulse)); mj_lambdas[self.mj_lambda2 as usize] = mj_lambda2; } pub fn solve2( &mut self, mj_lambdas: &mut [DeltaVel], mj_lambdas_pos: &mut [DeltaVel], ) { let mut mj_lambda2 = mj_lambdas[self.mj_lambda2 as usize]; let mut mj_lambda_pos2 = mj_lambdas_pos[self.mj_lambda2 as usize]; /* * Solve velocities. */ let ang_vel2 = self.ii2_sqrt.transform_vector(mj_lambda2.angular); let dlinvel = mj_lambda2.linear + ang_vel2.gcross(self.r2); let dangvel = ang_vel2; #[cfg(feature = "dim2")] let rhs = Vector3::new(dlinvel.x, dlinvel.y, dangvel) + self.rhs; #[cfg(feature = "dim3")] let dvel = Vector6::new( dlinvel.x, dlinvel.y, dlinvel.z, dangvel.x, dangvel.y, dangvel.z, ) + self.rhs; let new_impulse = (self.impulse + self.inv_lhs * dvel) .sup(&self.max_negative_impulse) .inf(&self.max_positive_impulse); let effective_impulse = new_impulse - self.impulse; self.impulse = new_impulse; let lin_impulse = effective_impulse.fixed_rows::(0).into_owned(); #[cfg(feature = "dim2")] let ang_impulse = effective_impulse[2]; #[cfg(feature = "dim3")] let ang_impulse = effective_impulse.fixed_rows::(3).into_owned(); mj_lambda2.linear -= self.im2 * lin_impulse; mj_lambda2.angular -= self .ii2_sqrt .transform_vector(ang_impulse + self.r2.gcross(lin_impulse)); /* * Solve positions. */ let ang_pos2 = self.ii2_sqrt.transform_vector(mj_lambda_pos2.angular); let dlinpos = mj_lambda_pos2.linear + ang_pos2.gcross(self.r2); let dangpos = ang_pos2; #[cfg(feature = "dim2")] let rhs = Vector3::new(dlinpos.x, dlinpos.y, dangpos) + self.rhs; #[cfg(feature = "dim3")] let dpos = Vector6::new( dlinpos.x, dlinpos.y, dlinpos.z, dangpos.x, dangpos.y, dangpos.z, ) + self.pos_rhs; let new_impulse = self.pos_impulse + self.inv_lhs * dpos; let effective_impulse = new_impulse - self.pos_impulse; self.pos_impulse = new_impulse; let lin_impulse = effective_impulse.fixed_rows::(0).into_owned(); #[cfg(feature = "dim2")] let ang_impulse = effective_impulse[2]; #[cfg(feature = "dim3")] let ang_impulse = effective_impulse.fixed_rows::(3).into_owned(); mj_lambda_pos2.linear -= self.im2 * lin_impulse; mj_lambda_pos2.angular -= self .ii2_sqrt .transform_vector(ang_impulse + self.r2.gcross(lin_impulse)); mj_lambdas[self.mj_lambda2 as usize] = mj_lambda2; mj_lambdas_pos[self.mj_lambda2 as usize] = mj_lambda_pos2; } // FIXME: duplicated code with the non-ground constraint. pub fn writeback_impulses(&self, joints_all: &mut [JointGraphEdge]) { let joint = &mut joints_all[self.joint_id].weight; if let JointParams::GenericJoint(fixed) = &mut joint.params { fixed.impulse = self.impulse; } } }