[Csp-commits] svn:csp r2157 - in trunk/csp: bin csplib/numeric cspsim cspsim/test data/ui/scripts/windows data/xml/objects data/xml/vehicles/aircraft/f16 data/xml/vehicles/aircraft/m2k

[<sv...@ww...>]

Author: mkrose
Date: 2007-07-29 23:40:43 -0700 (Sun, 29 Jul 2007)
New Revision: 2157

Added:
   trunk/csp/cspsim/test/
   trunk/csp/cspsim/test/test_PhysicsModel.cpp
Removed:
   trunk/csp/cspsim/AircraftPhysicsModel.cpp
   trunk/csp/cspsim/AircraftPhysicsModel.h
   trunk/csp/cspsim/DynamicalSystem.cpp
   trunk/csp/cspsim/DynamicalSystem.h
Modified:
   trunk/csp/bin/test_objects
   trunk/csp/csplib/numeric/NumericalMethod.cpp
   trunk/csp/csplib/numeric/NumericalMethod.h
   trunk/csp/csplib/numeric/Vector.h
   trunk/csp/csplib/numeric/VectorField.h
   trunk/csp/cspsim/BaseDynamics.cpp
   trunk/csp/cspsim/BaseDynamics.h
   trunk/csp/cspsim/CSPSim.cpp
   trunk/csp/cspsim/Collision.cpp
   trunk/csp/cspsim/Collision.h
   trunk/csp/cspsim/Engine.cpp
   trunk/csp/cspsim/FlightDynamics.cpp
   trunk/csp/cspsim/LandingGear.cpp
   trunk/csp/cspsim/LandingGear.h
   trunk/csp/cspsim/PhysicsModel.cpp
   trunk/csp/cspsim/PhysicsModel.h
   trunk/csp/cspsim/SConscript
   trunk/csp/cspsim/System.h
   trunk/csp/cspsim/SystemsModel.h
   trunk/csp/data/ui/scripts/windows/mainmenu.py
   trunk/csp/data/xml/objects/freefall_systems.xml
   trunk/csp/data/xml/vehicles/aircraft/f16/agent.xml
   trunk/csp/data/xml/vehicles/aircraft/f16/systems.xml
   trunk/csp/data/xml/vehicles/aircraft/m2k/agent.xml
   trunk/csp/data/xml/vehicles/aircraft/m2k/systems.xml
Log:
Update, simplify, and fix various minor issues with the aircraft
physics model.  This change eliminates DynamicalSystem and
AircraftPhysicsModel, making PhysicsModel the physics engine for
all vehicles.  Quite a few other changes were made as well:

* merge DynamicalSystem and AircraftPhysicsModel into PhysicsModel,
which should work for all vehicles.

* rewrite the physics solution in terms of position, velocity,
angular momentum, and attitude, where position and velocity
are in the global coordinate frame and angular momentum is
in the body frame.  angular momentum is initialized from angular
velocity at the start of each timestep and is not exposed
outside of PhysicsModel.  this corrects errors in the treatment
of the position integration, which was previously done in body
coordinates.  the correction is subtle under most circumstances
and probably not noticeable in the simulation.  tests have been
added to help verify the physics model and prevent regressions.

* eliminate custom valarray implementation (which I *think* was
not being used on any platform).  valarray itself could be
replaced by a simpler class since this change also unrolls all
of the valarray expressions to improve performance, at a small
cost to readability.

* add toBody and fromBody helper methods to BaseDynamics to make
coordinate conversions clearer in several dynamics routines.

* rename positionlocal and velocitylocal to position and velocity,
respectively.  the name "PositionLocal" is now used internally
by PhysicsModel to represent motion of the center of mass relative
to its position at the start of the timestep.

* in the gear simulation, update height, ground normal in body
coordinates, and wind velocity in body coordinates at each call
to computeForceAndMoment rather than only at preSimulationStep.
previously these values neglected aircraft motion during the
integration, leading to discontinuities in the response between
frames.  on the downside, the f16 aircraft is noticeably less
stable on landing, rocking back and forth between the main gear.
not sure yet which update caused this change, but it needs to
be investigated and (most likely) fixed.

* workaround uninitialized contact point of the gear simulation
during the first frame.

* fix possible NaN's in gear simulation due to incorrect inequality
tests.

* fix sim time at high frame rates (time was running too fast
when the frame rate exceeded 100 Hz).

* increase minimum frame rate from 10 Hz to 20 Hz

* various tweaks to GroundCollisionDynamics to try to improve
stability and make the response on impact less unrealistic
(in the absense of damage modeling).

* increase the initial height of the test aircraft so that they
settle more gracefully onto the runway.

Browse at: https://www.zerobar.net/viewcvs/viewcvs.cgi?view=rev&rev=2157

Diff omitted (133336 bytes).