Menu
▾
▴
[Csp-commits] APPLICATIONS/CSPSim CHANGES.current,1.33,1.34
|
From: <mk...@us...> - 2003-06-26 09:29:18
|
Update of /cvsroot/csp/APPLICATIONS/CSPSim
In directory sc8-pr-cvs1:/tmp/cvs-serv11273
Modified Files:
CHANGES.current
Log Message:
Index: CHANGES.current
===================================================================
RCS file: /cvsroot/csp/APPLICATIONS/CSPSim/CHANGES.current,v
retrieving revision 1.33
retrieving revision 1.34
diff -C2 -d -r1.33 -r1.34
*** CHANGES.current 22 Jun 2003 08:59:29 -0000 1.33
--- CHANGES.current 26 Jun 2003 09:29:15 -0000 1.34
***************
*** 2,5 ****
--- 2,80 ----
===========================
+ 2003-06-24: onsight
+ Added Profile.h which has simple but handy profiling
+ macros. There's a new PROFILE log target in Log.h as
+ well.
+
+ CSPSim.py now takes a --log parameter to specify which
+ log targets to record. E.g.:
+
+ CSPSim.py --log=PROFILE,PHYSICS
+
+ Note that if multiple targets are specified they must
+ be separated by commas (,) with no spaces. The target
+ names are the same as in Log.h, but without the leading
+ 'CSP_'.
+
+ Changed EngineDynamics to only reevaluate the thrust
+ during the preSimulationStep() call, rather than during
+ update(). This should still provide more than enough time
+ resolution.
+
+ Reimplemented the ground collision code to simplify the
+ design and fix some bugs.
+
+ Reimplemented the landing gear dynamics code. The design
+ is somewhat saner now, and most importantly it does not
+ make internal state changes during the rk evaluations.
+ The monolithic simulation method is now broken into more
+ manageable pieces, and some new features such as brake
+ temperature modeling have been added. There is still
+ some numerical instability when the wheels are locked and
+ the system behaves as a stiff harmonic oscillator (tires).
+ When the damping is too high, extreme instabilities can
+ occur. Even without violent behavior, the rk code generally
+ slows down and works too hard trying to simulate the
+ oscillation. Not sure at this point how best to fix the
+ problem.
+
+ Updated the GCC Makefile.in. Probably more changes still
+ needed to integrate CSPChunkLod smoothly.
+
+ ==========> WIN: Changed *PhysicModel to *PhysicsModel. Update the
+ project files accordingly:
+ PhysicsModel.cpp
+ AircraftPhysicsModel.cpp
+
+ ==========> WIN: Theater code now in separate subdirectories under
+ Include/Theater and Source/Theater. All the files in
+ Source/Theater need to be added to the project.
+
+ Changed BaseDynamics::update(dt) to computeForceAndMoment(x)
+ to reduce confusion with other update methods in the sim
+ that take real (sequential) time interval parameters.
+
+ 2003-06-22: onsight
+ Added StateSet to text objects to disable texturing with
+ tex units 1-3. Shouldn't be necessary, but it helps to keep
+ the text from disappearing when terrain multitexturing is
+ enabled. This is just a bandaid, not a cure.
+
+ Implemented brake_limit for landing gear. This is the
+ maximum static force (in newtons) that each brake can produce
+ (independent of tire friction). It should be set to zero
+ for wheels that do not have brakes.
+
+ Fixed a bug in the brake lowpass filter that reduced the
+ brake setting by 1/2.
+
+ Added brake-steering linkage, which allows much tighter
+ turning radii. Set brake_steering_linkage to -1.0 for
+ left-side gear and +1.0 for right-side gear. An additional
+ multiplier is passed as the second parameter to setSteering.
+ Brake-steering adds to normal braking, but the total
+ braking force is still constrained to [0, brake_limit] of
+ course.
+
2003-06-20: delta
Moved updateNearGround() outside loop in AircraftPhysicModel::doSimStep.
***************
*** 10,13 ****
--- 85,101 ----
added EngineDynamics::setThrottle.
+ 2003-06-19: onsight
+ Fixes in AircraftPhysicModel::doSimStep to correctly update
+ position when multiple internal timesteps are used.
+
+ Added missing Feature.cpp. Together with the postCreate
+ fix to SimData, makeFeatures is now running correctly.
+
+ The postCreate fix also cures the gear sprite placement
+ problem when starting in the air.
+
+ Renamed osgChunkLod-double director to CSPChunkLod, which
+ is now in cvs.
+
2003-06-19: delta
==========> GCC USERS: add Engine.cpp, BaseDynamics.cpp, Collision.cpp,
***************
*** 55,68 ****
--- 143,265 ----
~SmokeTrailSystem() and ~SceneModel().
+ 2003-06-11: onsight
+ Added a link from Feature to the associated FeatureObjectModel.
+
+ Added an "elevation correction" option to ObjectModel. The
+ default is true, which means that features will be floated
+ relative to the local terrain elevation. If false, Z=0 is
+ the reference elevation for the model. This can be used for
+ models that are decaled to the terrain surface and already
+ incorporate the local elevation variations. In principle,
+ this functionality and some other parts (e.g. contacts) should
+ be separated into subclases of ObjectModel (VehicleObjectModel,
+ DecalObjectModel, FeatureObjectModel, etc).
+
+ 2003-05-14: onsight
+ Some battlefield code cleanup.
+
+ New --log=XXX:YYY option for CSPSim to set the logging
+ classes. For example: --log=APP:BATTLEFIELD:PROFILE
+
+ Use for the new SimData LLA class to specify object coordinates
+ in latitude, longitude, altitude. TerrainObject now defines
+ a center latitude and longitude (set in XML), and the geographic
+ coordinates around this point are mapped into the sim's flat
+ coordinate space using a secant gnomonic projection. This
+ projection is not the same as was used for the Balkan terrain,
+ so object coordinates will not exactly match the terrain.
+ Internally, the (x, y) coordinate system is now relative to the
+ center of the terrain rather than the southwestern corner. The
+ internal coordinates used by demeter are unchanged, so an
+ offset is applied whenever camera positions are passed to
+ demeter.
+
+ 2003-05-06: onsight
+ New RandomForestModel for building random assortments of
+ 'X' quads. I haven't been able to turn of back-face culling
+ (not sure the normals would work out right even if I could),
+ so I'm drawing 4 quads per tree (both sides of each plane of
+ the 'X'). Seems to work, and is faster than billboarding,
+ but doesn't look as nice especially wrt lighting.
+
+ Added normals to billboard quads, renamed the class to
+ FeatureQuad.
+
+ Changed the interactions of LayoutTransform, FeatureSceneModel,
+ and ElevationCorrection. Fixed the elevation correction.
+
+ Renamed m_fdisToObject and m_fangleRot* to m_DistanceToObject
+ and m_AngleRot* in GameScreen. Added m_MinimumDistance which
+ is set to the active object's bounding sphere radius plus a
+ couple meters. The prevents clipping of the model by the
+ near plane in external views. You can still clip into other
+ models that are nearby of course.
+
+ Added a second SceneView to VirtualScene. The original SceneView
+ is now called m_FarView, and the new one is m_NearView. NearView
+ has the lighting group, and a group for objects near the camera.
+ NearView is only used if this later group is not empty.
+ VirtualScene::draw() first renders the FarView with the near
+ clip plane set to 1-3 m. This relatively large value helps to
+ reduce Z-fighting far from the camera. Then the depth buffer is
+ cleared and NearView is rendered with the depth frustum from 1
+ cm to 100 m. Typically the player's vehicle and interface is
+ rendered in the NearView for internal views. GameScreen manages
+ whether the NearView or FarView is used based on the current
+ viewing mode (internal or external).
+
+ Additional fixes to the HID keystate tracker.
+
+ 2003-05-05: onsight
+ Added HID::translate to keep track of the keyboard modifier
+ state for each key that is pressed, and force that modifier
+ state again when the key is released. So for example, if
+ PRESS{NUM-6} maps to "begin pan-right", and RELEASE{NUM-6}
+ maps to "end pan-right", then:
+
+ PRESS{NUM-6}, PRESS{SHIFT}, RELEASE{NUM-6}, RELEASE{SHIFT}
+
+ never sent the "end pan-right" message since the modifier state
+ when RELEASE{NUM-6} arrived included the shift state. The new
+ translate method records that PRESS{NUM-6} occurred with no
+ modifiers, and temporarily overrides the modifier state when
+ the key is released. This method is currently only called by
+ CSPSim::doInput(). It is called for all input events, and may
+ eventually find uses for other event types (e.g. joystick).
+
+ 2003-05-04: onsight
+ Small fix to the camera positioning correction that prevents it
+ from getting "stuck" to the ground.
+
+ Added FeatureQuad class to Terrain.h for creating textured quads
+ used in billboards (e.g. trees).
+
+ Similified RandomLayoutModel to use FeatureQuad.
+
+ Added a couple XML files for trees under Data/XML/theater/trees
+
+ Moved the osgDB search path setup out of VirtualScene and into
+ CSPSim::init(). All the paths are now relative to DataPath,
+ and there is a new accessor function in Config.h to make it
+ easier to access them. For example, getDataPath("ImagePath")
+ will return the full path for images (DataPath + ImagePath, if
+ ImagePath is not absolute).
+
2003-05-04: delta
Added a naive control of camera such that it keeps above ground.
+
Updated .net project.
+
Added a /Zm 1000 parameter in .net projects to avoid fatal
compiler error.
+
Changed vehicle orientation at start.
+
Changed some Atmosphere function members to return float instead
double to avoid some warnings.
+
Updated CSPSim.py to take account set_output <-> setOuput for
log file.
+
Added a vehicle to load in testing item in ini file. The vehicle
is specified in Vehicle.
|
