TODO - Integration has room for improvement

[shadowmage45]

• Currently using basic euler integration. Need to swap to some derivative based integration.
  • Pass in the vessel mass and COM into the integration code, use that to extrapolate positions for derivatives/substeps
    • treat entire craft as a single rigidbody for interpolated/extrapolated states, it should be at least close enough to correct for the duration being simulated, especially with the improvements to the base integration code
  • --how to handle changes in suspension compression from attitude changes or terrain changes that would happen in the substeps?
    • assume a single plane exists with surface-normal of the base raycast hit normal, use that plane for suspension interpolation data
  • -- how to accomodate changes to integration across multiple wheels on the same craft?
    • have the integrator simulate at the vessel level as for as rigidbody sub-integration
    • requires the wheel-collider integration code be aware of 'wheel body groups' and have customizable rigidbody com, mass, and ternsor data
    • could result in a fairly accurate simulation if handled properly, with potentially very little overhead cost (and would be much more efficient than if done on a per-wheel basis even at the worst case)

[shadowmage45]

Even without the ability to rework the physics, it seems that the current setup could still benefit from some form of multi-step integration for the spring and friction calculations.

• Given two states, at distinct points in time (A, B),
  • Two measurable values at each point - compression distance, surface position
  • Two deduced values - compression velocity, surface velocity
• The spring/damper and friction forces should represent the sum of forces across the interval in time as if it were not quantized, not the forces at a single point in time.
• Unsure on how to implement.
  • A simple 50/50 interpolation might be slightly more accurate, but seems like it would only add 1/2 frame of delay to everything; this could easily be accomplished by caching the last frames output data, combining with current frames calculations.
    • The 'delay' might actually be a good thing as it may smooth out acceleration/grip problems.