1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759
|
#ifndef SimTK_SIMBODY_RIGID_BODY_NODE_SPEC_H_
#define SimTK_SIMBODY_RIGID_BODY_NODE_SPEC_H_
/* -------------------------------------------------------------------------- *
* Simbody(tm) *
* -------------------------------------------------------------------------- *
* This is part of the SimTK biosimulation toolkit originating from *
* Simbios, the NIH National Center for Physics-Based Simulation of *
* Biological Structures at Stanford, funded under the NIH Roadmap for *
* Medical Research, grant U54 GM072970. See https://simtk.org/home/simbody. *
* *
* Portions copyright (c) 2005-12 Stanford University and the Authors. *
* Authors: Michael Sherman *
* Contributors: *
* Charles Schwieters (NIH): wrote the public domain IVM code from which *
* this was derived. *
* Peter Eastman: wrote the Euler Angle<->Quaternion conversion *
* *
* Licensed under the Apache License, Version 2.0 (the "License"); you may *
* not use this file except in compliance with the License. You may obtain a *
* copy of the License at http://www.apache.org/licenses/LICENSE-2.0. *
* *
* Unless required by applicable law or agreed to in writing, software *
* distributed under the License is distributed on an "AS IS" BASIS, *
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. *
* See the License for the specific language governing permissions and *
* limitations under the License. *
* -------------------------------------------------------------------------- */
/**@file
* This file contains just the templatized class RigidBodyNodeSpec<dof> which
* is used as the base class for most of the RigidBodyNodes (that is, the
* implementations of Mobilizers). The only exceptions are nodes whose
* mobilizers provide no degrees of freedom -- Ground and Weld.
*
* This file contains all the multibody mechanics method declarations that
* involve a single body and its mobilizer (inboard joint), that is, one node
* in the multibody tree. These methods constitute the inner loops of the
* multibody calculations, and much suffering is undergone here to make them
* run fast. In particular most calculations are templatized by the number of
* mobilities, so that compile-time sizes are known for everything.
*
* Most methods here expect to be called in a particular order during traversal
* of the tree -- either base to tip or tip to base.
*/
#include "SimbodyMatterSubsystemRep.h"
#include "RigidBodyNode.h"
/**
* This still-abstract class is a skeleton implementation of a built-in
* mobilizer, with the number of mobilities (dofs, u's) specified as a template
* parameter. That way all the code that is common except for the dimensionality
* of the mobilizer can be written once, and the compiler generates specific
* implementations for each of the six possible dimensionalities (1-6
* mobilities). Each implementation works only on fixed size Vec<> and Mat<>
* types, so can use very high speed inline operators.
*/
template<int dof, bool noR_FM=false, bool noX_MB=false, bool noR_PF=false>
class RigidBodyNodeSpec : public RigidBodyNode {
public:
RigidBodyNodeSpec(const MassProperties& mProps_B,
const Transform& X_PF,
const Transform& X_BM,
UIndex& nextUSlot,
USquaredIndex& nextUSqSlot,
QIndex& nextQSlot,
QDotHandling qdotHandling,
QuaternionUse quaternionUse,
bool isReversed)
: RigidBodyNode(mProps_B, X_PF, X_BM,
qdotHandling, quaternionUse, isReversed)
{
// don't call any virtual methods in here!
uIndex = nextUSlot;
uSqIndex = nextUSqSlot;
qIndex = nextQSlot;
}
void updateSlots(UIndex& nextUSlot, USquaredIndex& nextUSqSlot, QIndex& nextQSlot) {
// OK to call virtual method here.
nextUSlot += getDOF();
nextUSqSlot += getDOF()*getDOF();
nextQSlot += getMaxNQ();
}
virtual ~RigidBodyNodeSpec() {}
// This is the type of the joint transition matrix H, but our definition
// of H is transposed from Jain's or Schwieters'. That is, what we're
// calling H here is what Jain calls H* and Schwieters calls H^T. So
// our H matrix is 6 x nu, or more usefully it is 2 rows of Vec3's.
// The first row define how u's contribute to angular velocities; the
// second defines how u's contribute to linear velocities.
typedef Mat<2,dof,Vec3> HType;
// Provide default implementations for setQToFitTransformImpl() and setQToFitVelocityImpl()
// which are implemented using the rotational and translational quantity routines. These assume
// that the rotational and translational coordinates are independent, with rotation handled
// first and then left alone. If a mobilizer type needs to deal with rotation and
// translation simultaneously, it should provide a specific implementation for these two routines.
// *Each* mobilizer must implement setQToFit{Rotation,Translation} and
// setUToFit{AngularVelocity,LinearVelocity}; there are no defaults.
virtual void setQToFitTransformImpl(const SBStateDigest& sbs, const Transform& X_FM, Vector& q) const {
setQToFitRotationImpl (sbs,X_FM.R(),q);
setQToFitTranslationImpl(sbs,X_FM.p(),q);
}
virtual void setUToFitVelocityImpl(const SBStateDigest& sbs, const Vector& q, const SpatialVec& V_FM, Vector& u) const {
setUToFitAngularVelocityImpl(sbs,q,V_FM[0],u);
setUToFitLinearVelocityImpl (sbs,q,V_FM[1],u);
}
// The following routines calculate joint-specific position kinematic
// quantities. They may assume that *all* position kinematics (not just
// joint-specific) has been done for the parent, and that the position
// state variables q are available. Each routine may also assume that the
// previous routines have been called, in the order below.
// The routines are structured as operators -- they use the State but
// do not change anything in it, including the cache. Instead they are
// passed arguments to write their results into. In practice, these
// arguments will typically be in the State cache (see below).
// This mandatory routine calculates the joint transition matrix H_FM, giving
// the change of velocity induced by the generalized speeds u for this
// mobilizer, expressed in the mobilizer's inboard "fixed" frame F (attached to
// this body's parent).
// This method constitutes the *definition* of the generalized speeds for
// a particular joint.
// This routine can depend on X_FM having already
// been calculated and available in the PositionCache but must NOT depend
// on any quantities involving Ground or other bodies.
// Note: this calculates the H matrix *as defined*; we might reverse
// the frames in use. So we call this H_F0M0 while the possibly reversed
// version is H_FM. Be sure to use X_F0M0 in your calculation for H_F0M0
// if you need access to the cross-mobilizer transform.
// CAUTION: our definition of H is transposed from Jain and Schwieters.
virtual void calcAcrossJointVelocityJacobian(
const SBStateDigest& sbs,
HType& H_F0M0) const=0;
// This mandatory routine calculates the time derivative taken in F of
// the matrix H_FM (call it HDot_FM). This is zero if the generalized
// speeds are all defined in terms of the F frame, which is often the case.
// This routine can depend on X_FM and H_FM being available already in
// the state PositionCache, and V_FM being already in the state VelocityCache.
// However, it must NOT depend on any quantities involving Ground or other bodies.
// Note: this calculates the HDot matrix *as defined*; we might reverse
// the frames in use. So we call this HDot_F0M0 while the possibly reversed
// version is HDot_FM. Be sure to use X_F0M0 and V_F0M0 in your calculation
// of HDot_F0M0 if you need access to the cross-mobilizer transform and/or velocity.
// CAUTION: our definition of H is transposed from Jain and Schwieters.
virtual void calcAcrossJointVelocityJacobianDot(
const SBStateDigest& sbs,
HType& HDot_F0M0) const = 0;
// We allow a mobilizer to be defined in reverse when that is more
// convenient. That is, the H matrix can be defined by giving H_F0M0=H_MF and
// HDot_F0M0=HDot_MF instead of H_FM and HDot_FM. In that case the
// following methods are called instead of the above; the default
// implementation postprocesses the output from the above methods, but a
// mobilizer can override it if it knows how to get the job done faster.
virtual void calcReverseMobilizerH_FM(
const SBStateDigest& sbs,
HType& H_FM) const;
virtual void calcReverseMobilizerHDot_FM(
const SBStateDigest& sbs,
HType& HDot_FM) const;
// This routine is NOT joint specific, but cannot be called until the across-joint
// transform X_FM has been calculated and is available in the State cache.
void calcBodyTransforms(
const SBTreePositionCache& pc,
Transform& X_PB,
Transform& X_GB) const
{
const Transform& X_MB = getX_MB(); // fixed
const Transform& X_PF = getX_PF(); // fixed
const Transform& X_FM = getX_FM(pc); // just calculated
const Transform& X_GP = getX_GP(pc); // already calculated
const Transform X_FB = (noX_MB ? X_FM : X_FM*X_MB); // 63 flops
X_PB = (noR_PF ? Transform(X_FB.R(), X_PF.p()+X_FB.p()) : X_PF*X_FB); // 63 flops
X_GB = X_GP * X_PB; // 63 flops
}
// Same for all mobilizers. The return matrix here is precisely the
// one used by Jain and Schwieters, but transposed.
void calcParentToChildVelocityJacobianInGround(
const SBModelVars& mv,
const SBTreePositionCache& pc,
HType& H_PB_G) const;
// Same for all mobilizers. This is the time derivative of
// the matrix H_PB_G above, with the derivative taken in the
// Ground frame.
void calcParentToChildVelocityJacobianInGroundDot(
const SBModelVars& mv,
const SBTreePositionCache& pc,
const SBTreeVelocityCache& vc,
HType& HDot_PB_G) const;
void realizeModel(SBStateDigest& sbs) const
{
}
void realizeInstance(const SBStateDigest& sbs) const
{
}
// Set a new configuration and calculate the consequent kinematics.
// Must call base-to-tip.
void realizePosition(const SBStateDigest& sbs) const
{
const SBModelVars& mv = sbs.getModelVars();
const SBModelCache& mc = sbs.getModelCache();
const SBInstanceCache& ic = sbs.getInstanceCache();
const Vector& allQ = sbs.getQ();
SBTreePositionCache& pc = sbs.updTreePositionCache();
Vector& allQErr = sbs.updQErr();
// Mobilizer specific.
const SBModelPerMobodInfo& mbInfo = getModelInfo(mc);
// First perform precalculations on these new q's, such as stashing away
// sines and cosines of angles. We'll put all the results into the State's
// position cache for later use. If there are local constraints on the q's
// (currently that means quaternion normalization constraints), then we
// calculate those here and put the result in the appropriate slot of qerr.
const int nq=mbInfo.nQInUse, nqpool=mbInfo.nQPoolInUse,
nqerr=(mbInfo.hasQuaternionInUse ? 1 : 0);
const Real* q0 = nq ? &allQ[mbInfo.firstQIndex] : 0;
Real* qpool0= nqpool ? &pc.mobilizerQCache[mbInfo.startInQPool] : 0;
Real* qerr0 = nqerr ? &allQErr[ic.firstQuaternionQErrSlot
+ mbInfo.quaternionPoolIndex]
: 0;
performQPrecalculations(sbs, q0, nq, qpool0, nqpool, qerr0, nqerr);
// Now that we've done the necessary precalculations, calculate the cross-
// mobilizer transform X_FM without recalculating anything. For reversed
// mobilizers we have to do our own reversal here because mobilizers don't
// handle that themselves.
if (isReversed()) {
Transform X_MF;
calcX_FM(sbs, q0, nq, qpool0, nqpool, X_MF);
updX_FM(pc) = ~X_MF;
} else
calcX_FM(sbs, q0, nq, qpool0, nqpool, updX_FM(pc));
// With X_FM in the cache, and X_GP for the parent already calculated (we're doing
// an outward pass), we can calculate X_PB and X_GB now.
calcBodyTransforms(pc, updX_PB(pc), updX_GB(pc));
// Here we do allow the mobilizer to calculate the reversed H matrix, but the
// default implementation of the reversed method just calls the forward method
// and then reverses it. For some mobilizers that is unreasonably expensive.
// REMINDER: our H matrix definition is transposed from Jain and Schwieters.
if (isReversed()) calcReverseMobilizerH_FM (sbs, updH_FM(pc));
else calcAcrossJointVelocityJacobian(sbs, updH_FM(pc));
// Here we're using the cross-mobilizer hinge matrix H_FM that we just
// calculated to compute H(==H_PB_G), the equivalent hinge matrix between
// the parent's body frame P and child's body frame B, but expressed in
// Ground. (F is fixed on P and M is fixed on B.)
calcParentToChildVelocityJacobianInGround(mv,pc, updH(pc));
// Mobilizer independent.
calcJointIndependentKinematicsPos(pc);
}
// Set new velocities for the current configuration, and calculate
// all the velocity-dependent terms. Must call base-to-tip.
// This routine may assume that *all* position
// kinematics (not just joint-specific) has been done for this node,
// that all velocity kinematics has been done for the parent, and
// that the velocity state variables (u) are available. The
// quantities that must be computed are:
// V_FM relative velocity of B's M frame in P's F frame,
// expressed in F (note: this is also V_PM_F since
// F is fixed on P). (This is H_FM*u.)
// V_PB_G relative velocity of B in P (==H*u), expr. in G
// HDot_FM time derivative of hinge matrix H_FM
// HDot time derivative of H (==H_PB_G) hinge matrix relating body frames
// VD_PB_G acceleration remainder term HDot*u, expr. in G
// The code is the same for all joints, although parametrized by ndof.
void realizeVelocity(const SBStateDigest& sbs) const
{
const SBModelVars& mv = sbs.getModelVars();
const SBTreePositionCache& pc = sbs.getTreePositionCache();
SBTreeVelocityCache& vc = sbs.updTreeVelocityCache();
const Vector& allU = sbs.getU();
const Vec<dof>& u = fromU(allU);
// Mobilizer specific.
calcQDot(sbs, &allU[uIndex], &sbs.updQDot()[qIndex]);
updV_FM(vc) = getH_FM(pc) * u; // 6*dof flops
updV_PB_G(vc) = getH(pc) * u; // 6*dof flops
// REMINDER: our H matrix definition is transposed from Jain and Schwieters.
if (isReversed()) calcReverseMobilizerHDot_FM (sbs, updHDot_FM(vc));
else calcAcrossJointVelocityJacobianDot(sbs, updHDot_FM(vc));
// Here we're using the cross-mobilizer hinge matrix derivative HDot_FM
// that we just calculated to compute HDot(==HDot_PB_G), the derivative
// of H (==H_PB_G) between the parent's body frame P and child's body
// frame B. The derivative is taken in the Ground frame and the result
// is expressed in Ground. (F is fixed on P and M is fixed on B.)
calcParentToChildVelocityJacobianInGroundDot(mv,pc,vc, updHDot(vc));
updVD_PB_G(vc) = getHDot(vc) * u; // 6*dof flops
// Mobilizer independent.
calcJointIndependentKinematicsVel(pc,vc);
}
// Articulated body inertias have been calculated; here we're
void realizeDynamics(const SBArticulatedBodyInertiaCache& abc,
const SBStateDigest& sbs) const
{
const SBTreePositionCache& pc = sbs.getTreePositionCache();
const SBTreeVelocityCache& vc = sbs.getTreeVelocityCache();
SBDynamicsCache& dc = sbs.updDynamicsCache();
// Mobilizer-specific.
// None.
// Mobilizer independent.
calcJointIndependentDynamicsVel(pc,abc,vc,dc);
}
// There is no realizeAcceleration().
void realizeReport(const SBStateDigest& sbs) const
{
}
// Use base class implementation of calcCompositeBodyInertiasInward() since
// that is independent of mobilities.
// This is a dynamics-stage calculation and must be called tip-to-base (inward).
void realizeArticulatedBodyInertiasInward(
const SBInstanceCache& ic,
const SBTreePositionCache& pc,
SBArticulatedBodyInertiaCache& abc) const;
// calcJointIndependentDynamicsVel() must be called after ArticulatedBodyInertias.
// This dynamics-stage calculation is needed for handling constraints. It
// must be called base-to-tip (outward);
void realizeYOutward(
const SBInstanceCache& ic,
const SBTreePositionCache& pc,
const SBArticulatedBodyInertiaCache& abc,
SBDynamicsCache& dc) const;
// These routines give each node a chance to set appropriate defaults in a piece
// of the state corresponding to a particular stage. Default implementations here
// assume non-ball joint; override if necessary.
virtual void setMobilizerDefaultModelValues (const SBTopologyCache&, SBModelVars&) const {}
virtual void setMobilizerDefaultInstanceValues(const SBModelVars&, SBInstanceVars&) const {}
virtual void setMobilizerDefaultTimeValues (const SBModelVars&, SBTimeVars&) const {}
virtual void setMobilizerDefaultPositionValues(const SBModelVars& s, Vector& q) const
{
toQ(q) = 0;
}
virtual void setMobilizerDefaultVelocityValues(const SBModelVars&, Vector& u) const
{
toU(u) = 0;
}
virtual void setMobilizerDefaultDynamicsValues(const SBModelVars&, SBDynamicsVars&) const {}
virtual void setMobilizerDefaultAccelerationValues(const SBModelVars&,
SBDynamicsVars& v) const {}
int getDOF() const {return dof;}
virtual int getMaxNQ() const {
assert(quaternionUse == QuaternionIsNeverUsed);
return dof; // maxNQ can be larger than dof if there's a quaternion
}
// Default method must be overridden if there is a chance that NQ
// might not be the same as NU.
virtual int getNQInUse(const SBModelVars&) const {
assert(quaternionUse == QuaternionIsNeverUsed);
return dof; // DOF <= NQ <= maxNQ
}
// Currently NU is always just the Mobilizer's DOFs (the template argument)
// Later we may want to offer modeling options to lock joints, or perhaps
// break them.
virtual int getNUInUse(const SBModelVars&) const {
return dof;
}
// Default method must be overridden if this mobilizer might use a
// quaternion under some conditions.
virtual bool isUsingQuaternion(const SBStateDigest&, MobilizerQIndex& startOfQuaternion) const {
assert(quaternionUse == QuaternionIsNeverUsed);
startOfQuaternion.invalidate();
return false;
}
// You must override the next few methods if qdot might not be the same
// as u under *any* circumstance.
// This method should calculate qdot=N*u, where N=N(q) is the kinematic
// coupling matrix. State digest should be at Stage::Position.
virtual void calcQDot(const SBStateDigest&,
const Real* u, Real* qdot) const
{
assert(qdotHandling == QDotIsAlwaysTheSameAsU);
Vec<dof>::updAs(qdot) = Vec<dof>::getAs(u); // default says qdot=u
}
// This method should calculate u=N^-1*qdot, where N=N(q).
// State digest should be at Stage::Position.
virtual void calcLocalUFromLocalQDot(const SBStateDigest&,
const Real* qdot, Real* u) const
{
assert(qdotHandling == QDotIsAlwaysTheSameAsU);
Vec<dof>::updAs(u) = Vec<dof>::getAs(qdot); // default says u=qdot
}
// This method should calculate qdotdot=N*udot + NDot*u, where N=N(q),
// NDot=NDot(q,u). State digest should be at Stage::Velocity.
virtual void calcQDotDot(const SBStateDigest&,
const Real* udot, Real* qdotdot) const
{
assert(qdotHandling == QDotIsAlwaysTheSameAsU);
Vec<dof>::updAs(qdotdot) = Vec<dof>::getAs(udot); // default: qdotdot=udot
}
// This method should calculate udot=N^-1*(qdotdot-NDot*u), where N=N(q),
// NDot=NDot(q,u). State digest should be at Stage::Velocity.
virtual void calcLocalUDotFromLocalQDotDot(const SBStateDigest&,
const Real* qdotdot, Real* udot) const
{
assert(qdotHandling == QDotIsAlwaysTheSameAsU);
Vec<dof>::updAs(udot) = Vec<dof>::getAs(qdotdot); // default: udot=qdotdot
}
// State digest should be at Stage::Position for calculating N (the matrix that
// maps generalized speeds u to coordinate derivatives qdot, qdot=Nu). The
// default implementation assumes nq==nu and the nqXnu block of N corresponding
// to this mobilizer is identity. Then either operation (regardless of side)
// just copies nu numbers from in to out.
//
// THIS MUST BE OVERRIDDEN by any mobilizer for which nq != nu, or qdot != u.
virtual void multiplyByN(const SBStateDigest&, bool matrixOnRight,
const Real* in, Real* out) const
{
assert(qdotHandling == QDotIsAlwaysTheSameAsU);
Vec<dof>::updAs(out) = Vec<dof>::getAs(in);
}
virtual void multiplyByNInv(const SBStateDigest&, bool matrixOnRight,
const Real* in, Real* out) const
{
assert(qdotHandling == QDotIsAlwaysTheSameAsU);
Vec<dof>::updAs(out) = Vec<dof>::getAs(in);
}
// State digest should be at Stage::Velocity for calculating NDot (the matrix
// that is used in mapping generalized accelerations udot to coordinate 2nd
// derivatives qdotdot, qdotdot=N udot + NDot u. The default implementation
// assumes nq==nu and the nqXnu block of N corresponding to this mobilizer is
// identity, so NDot is an nuXnu block of zeroes. Then either operation
// (regardless of side) just copies nu zeros to out.
//
// THIS MUST BE OVERRIDDEN by any mobilizer for which nq != nu, or qdot != u.
virtual void multiplyByNDot(const SBStateDigest&, bool matrixOnRight,
const Real* in, Real* out) const
{
assert(qdotHandling == QDotIsAlwaysTheSameAsU);
Vec<dof>::updAs(out) = 0;
}
// No default implementations here for:
// calcMobilizerTransformFromQ
// calcMobilizerVelocityFromU
// calcMobilizerAccelerationFromUDot
// calcParentToChildTransformFromQ
// calcParentToChildVelocityFromU
// calcParentToChildAccelerationFromUDot
virtual void setVelFromSVel(
const SBTreePositionCache& pc,
const SBTreeVelocityCache& vc,
const SpatialVec& sVel,
Vector& u) const;
// Return true if any change is made to the q vector.
virtual bool enforceQuaternionConstraints(
const SBStateDigest& sbs,
Vector& q,
Vector& qErrest) const
{
assert(quaternionUse == QuaternionIsNeverUsed);
return false;
}
void convertToEulerAngles(const Vector& inputQ, Vector& outputQ) const {
// The default implementation just copies Q. Subclasses may override this.
assert(quaternionUse == QuaternionIsNeverUsed);
toQ(outputQ) = fromQ(inputQ);
}
void convertToQuaternions(const Vector& inputQ, Vector& outputQ) const {
// The default implementation just copies Q. Subclasses may override this.
assert(quaternionUse == QuaternionIsNeverUsed);
toQ(outputQ) = fromQ(inputQ);
}
// Get a column of H_PB_G, which is what Jain calls H* and Schwieters calls H^T.
const SpatialVec& getHCol(const SBTreePositionCache& pc, int j) const {
return getH(pc)(j);
}
// Get a column of H_FM the local cross-mobilizer hinge matrix expressed in the
// parent (inboard) mobilizer frame F.
const SpatialVec& getH_FMCol(const SBTreePositionCache& pc, int j) const {
return getH_FM(pc)(j);
}
// Access to body-oriented state and cache entries is the same for all nodes,
// and joint oriented access is almost the same but parametrized by dof. There is a special
// case for quaternions because they use an extra state variable, and although we don't
// have to we make special scalar routines available for 1-dof joints. Note that all State access
// routines are inline, not virtual, so the cost is just an indirection and an index.
//
// TODO: these inner-loop methods probably shouldn't be indexing a Vector, which requires
// several indirections. Instead, the top-level caller should find the Real* data contained in the
// Vector and then pass that to the RigidBodyNode routines which will call these ones.
// General joint-dependent select-my-goodies-from-the-pool routines.
const Vec<dof>& fromQ (const Real* q) const {return Vec<dof>::getAs(&q[qIndex]);}
Vec<dof>& toQ ( Real* q) const {return Vec<dof>::updAs(&q[qIndex]);}
const Vec<dof>& fromU (const Real* u) const {return Vec<dof>::getAs(&u[uIndex]);}
Vec<dof>& toU ( Real* u) const {return Vec<dof>::updAs(&u[uIndex]);}
const Mat<dof,dof>& fromUSq(const Real* uSq) const {return Mat<dof,dof>::getAs(&uSq[uSqIndex]);}
Mat<dof,dof>& toUSq ( Real* uSq) const {return Mat<dof,dof>::updAs(&uSq[uSqIndex]);}
// Same, but specialized for the common case where dof=1 so everything is scalar.
const Real& from1Q (const Real* q) const {return q[qIndex];}
Real& to1Q ( Real* q) const {return q[qIndex];}
const Real& from1U (const Real* u) const {return u[uIndex];}
Real& to1U ( Real* u) const {return u[uIndex];}
const Real& from1USq(const Real* uSq) const {return uSq[uSqIndex];}
Real& to1USq ( Real* uSq) const {return uSq[uSqIndex];}
// Same, specialized for quaternions. We're assuming that the quaternion is stored
// in the *first* four coordinates, if there are more than four altogether.
const Vec4& fromQuat(const Real* q) const {return Vec4::getAs(&q[qIndex]);}
Vec4& toQuat ( Real* q) const {return Vec4::updAs(&q[qIndex]);}
// Extract a Vec3 from a Q-like or U-like object, beginning at an offset from the qIndex or uIndex.
const Vec3& fromQVec3(const Real* q, int offs) const {return Vec3::getAs(&q[qIndex+offs]);}
Vec3& toQVec3 ( Real* q, int offs) const {return Vec3::updAs(&q[qIndex+offs]);}
const Vec3& fromUVec3(const Real* u, int offs) const {return Vec3::getAs(&u[uIndex+offs]);}
Vec3& toUVec3 ( Real* u, int offs) const {return Vec3::updAs(&u[uIndex+offs]);}
// These provide an identical interface for when q,u are given as Vectors rather than Reals.
const Vec<dof>& fromQ (const Vector& q) const {return fromQ(&q[0]);} // convert to array of Real
Vec<dof>& toQ ( Vector& q) const {return toQ (&q[0]);}
const Vec<dof>& fromU (const Vector& u) const {return fromU(&u[0]);}
Vec<dof>& toU ( Vector& u) const {return toU (&u[0]);}
const Mat<dof,dof>& fromUSq(const Vector& uSq) const {return fromUSq(&uSq[0]);}
Mat<dof,dof>& toUSq ( Vector& uSq) const {return toUSq (&uSq[0]);}
const Real& from1Q (const Vector& q) const {return from1Q (&q[0]);}
Real& to1Q ( Vector& q) const {return to1Q (&q[0]);}
const Real& from1U (const Vector& u) const {return from1U (&u[0]);}
Real& to1U ( Vector& u) const {return to1U (&u[0]);}
const Real& from1USq(const Vector& uSq) const {return from1USq(&uSq[0]);}
Real& to1USq ( Vector& uSq) const {return to1USq (&uSq[0]);}
// Same, specialized for quaternions. We're assuming that the quaternion comes first in the coordinates.
const Vec4& fromQuat(const Vector& q) const {return fromQuat(&q[0]);}
Vec4& toQuat ( Vector& q) const {return toQuat (&q[0]);}
// Extract a Vec3 from a Q-like or U-like object, beginning at an offset from the qIndex or uIndex.
const Vec3& fromQVec3(const Vector& q, int offs) const {return fromQVec3(&q[0], offs);}
Vec3& toQVec3 ( Vector& q, int offs) const {return toQVec3 (&q[0], offs);}
const Vec3& fromUVec3(const Vector& u, int offs) const {return fromUVec3(&u[0], offs);}
Vec3& toUVec3 ( Vector& u, int offs) const {return toUVec3 (&u[0], offs);}
// Applications of the above extraction routines to particular interesting items in the State. Note
// that you can't use these for quaternions since they extract "dof" items.
// Cache entries (cache is mutable in a const State)
// Position
// CAUTION: our H definition is transposed from Jain and Schwieters.
const HType& getH_FM(const SBTreePositionCache& pc) const
{ return HType::getAs(&pc.storageForH_FM[2*uIndex]); }
HType& updH_FM(SBTreePositionCache& pc) const
{ return HType::updAs(&pc.storageForH_FM[2*uIndex]); }
// "H" here should really be H_PB_G, that is, cross joint transition
// matrix relating parent and body frames, but expressed in Ground.
// CAUTION: our H definition is transposed from Jain and Schwieters.
const HType& getH(const SBTreePositionCache& pc) const
{ return HType::getAs(&pc.storageForH[2*uIndex]); }
HType& updH(SBTreePositionCache& pc) const
{ return HType::updAs(&pc.storageForH[2*uIndex]); }
// Velocity
// CAUTION: our H definition is transposed from Jain and Schwieters.
const HType& getHDot_FM(const SBTreeVelocityCache& vc) const
{ return HType::getAs(&vc.storageForHDot_FM[2*uIndex]); }
HType& updHDot_FM(SBTreeVelocityCache& vc) const
{ return HType::updAs(&vc.storageForHDot_FM[2*uIndex]); }
// CAUTION: our H definition is transposed from Jain and Schwieters.
const HType& getHDot(const SBTreeVelocityCache& vc) const
{ return HType::getAs(&vc.storageForHDot[2*uIndex]); }
HType& updHDot(SBTreeVelocityCache& vc) const
{ return HType::updAs(&vc.storageForHDot[2*uIndex]); }
// Dynamics
// These are calculated with articulated body inertias.
const Mat<dof,dof>& getD(const SBArticulatedBodyInertiaCache& abc) const
{return fromUSq(abc.storageForD);}
Mat<dof,dof>& updD(SBArticulatedBodyInertiaCache& abc) const
{return toUSq (abc.storageForD);}
const Mat<dof,dof>& getDI(const SBArticulatedBodyInertiaCache& abc) const
{return fromUSq(abc.storageForDI);}
Mat<dof,dof>& updDI(SBArticulatedBodyInertiaCache& abc) const
{return toUSq (abc.storageForDI);}
const Mat<2,dof,Vec3>& getG(const SBArticulatedBodyInertiaCache& abc) const
{ return Mat<2,dof,Vec3>::getAs(&abc.storageForG[2*uIndex]); }
Mat<2,dof,Vec3>& updG(SBArticulatedBodyInertiaCache& abc) const
{ return Mat<2,dof,Vec3>::updAs(&abc.storageForG[2*uIndex]); }
const Vec<dof>& getTau(const SBInstanceCache& ic, const Real* tau) const {
const PresForcePoolIndex tauIx = ic.getMobodInstanceInfo(nodeNum).firstPresForce;
assert(tauIx.isValid());
return Vec<dof>::getAs(&tau[tauIx]);
}
Vec<dof>& updTau(const SBInstanceCache& ic, Real* tau) const {
const PresForcePoolIndex tauIx = ic.getMobodInstanceInfo(nodeNum).firstPresForce;
assert(tauIx.isValid());
return Vec<dof>::updAs(&tau[tauIx]);
}
// Acceleration
const Vec<dof>& getEpsilon (const SBTreeAccelerationCache& ac) const {return fromU (ac.epsilon);}
Vec<dof>& updEpsilon (SBTreeAccelerationCache& ac) const {return toU (ac.epsilon);}
const Real& get1Epsilon(const SBTreeAccelerationCache& ac) const {return from1U(ac.epsilon);}
Real& upd1Epsilon(SBTreeAccelerationCache& ac) const {return to1U (ac.epsilon);}
void multiplyBySystemJacobian(
const SBTreePositionCache& pc,
const Real* v,
SpatialVec* Jv) const;
void multiplyBySystemJacobianTranspose(
const SBTreePositionCache& pc,
SpatialVec* zTmp,
const SpatialVec* X,
Real* JtX) const;
void calcEquivalentJointForces(
const SBTreePositionCache& pc,
const SBDynamicsCache& dc,
const SpatialVec* bodyForces,
SpatialVec* allZ,
Real* jointForces) const;
void calcUDotPass1Inward(
const SBInstanceCache& ic,
const SBTreePositionCache& pc,
const SBArticulatedBodyInertiaCache&,
const SBDynamicsCache& dc,
const Real* jointForces,
const SpatialVec* bodyForces,
const Real* allUDot,
SpatialVec* allZ,
SpatialVec* allGepsilon,
Real* allEpsilon) const;
void calcUDotPass2Outward(
const SBInstanceCache& ic,
const SBTreePositionCache& pc,
const SBArticulatedBodyInertiaCache&,
const SBTreeVelocityCache& vc,
const SBDynamicsCache& dc,
const Real* epsilonTmp,
SpatialVec* allA_GB,
Real* allUDot,
Real* allTau) const;
void multiplyByMInvPass1Inward(
const SBInstanceCache& ic,
const SBTreePositionCache& pc,
const SBArticulatedBodyInertiaCache&,
const Real* f,
SpatialVec* allZ,
SpatialVec* allGepsilon,
Real* allEpsilon) const OVERRIDE_11;
void multiplyByMInvPass2Outward(
const SBInstanceCache& ic,
const SBTreePositionCache& pc,
const SBArticulatedBodyInertiaCache&,
const Real* epsilonTmp,
SpatialVec* allA_GB,
Real* allUDot) const OVERRIDE_11;
// Also serves as pass 1 for inverse dynamics.
void calcBodyAccelerationsFromUdotOutward(
const SBTreePositionCache& pc,
const SBTreeVelocityCache& vc,
const Real* allUDot,
SpatialVec* allA_GB) const;
void calcInverseDynamicsPass2Inward(
const SBTreePositionCache& pc,
const SBTreeVelocityCache& vc,
const SpatialVec* allA_GB,
const Real* jointForces,
const SpatialVec* bodyForces,
SpatialVec* allFTmp,
Real* allTau) const;
void multiplyByMPass1Outward(
const SBTreePositionCache& pc,
const Real* allUDot,
SpatialVec* allA_GB) const;
void multiplyByMPass2Inward(
const SBTreePositionCache& pc,
const SpatialVec* allA_GB,
SpatialVec* allFTmp,
Real* allTau) const;
};
#endif // SimTK_SIMBODY_RIGID_BODY_NODE_SPEC_H_
|