tatlin
commented
12 years ago Works now. I've got it more stable with shorter spring rest length, tighter steps and more dampening. In example below there are 10 points laid out along x-axis every 100 feet. Spring rest length is 1 foot. You can see the contraction in the screenshot below. You can see from the forces in the m_Velocity field and the delta between m_oldPosition and m_Position that the point is moving to the left. What i'm not sure, because the dynamics are not working yet is how well the dampening works over time and does acceleration stay under control or not.
Regarding dynamics - I'm still experimenting with timers and having trouble having the stopwatch trigger an update when the interval is hit. I have recreated the old dynTimer and am using a stopwatch but the only way I can hit KeepTime() is in the Evaluate loop for dynTimer. I want to send a pulse downstream every time the stopwatch hits a specified interval in milliseconds. I feel a mental block coming on here - I need some Stephen magic or at least a hint from Stephen or Ian.
Todo on the dynDynamicRelaxation:
- parse input point and curve lists, convert to particles and springs. Right now it just uses a stub setup method in the constructor
- Make the output points and curves be handled non-destructively, this will speed things up a lot
- Manage changes to input lists (add,remove,change)
- Expose stepsize as an input port.
- Coloring of springs would fun but adding OSTs is a PIA in revit, maybe use AVF instead.
Steell
ekatzenstein
dynRelaxation node implements a basic particle system input - list of ref points get mapped to particles (use Revit XYZ class for now for Vector3d) input - list of line-bypoints mapped to springs input - timestep - seperate node to set ms, either implement a timer node or set ms as a double fixity - read parameter / data set on user-created ref points (bool fixed) gravity and forces - hard code gravity and drag for now in the PS. solver can handle wind as well but let's get the basics going first.
internals - solver - verlet or rk4 particle.cs spring.cs particlesystem.cs
dynRelaxation node is responsible for managing inputs, instantiating the PS, creating particles from ref points and springs from lines, and then calling the update as part of Evaluate. after PS update step, PS update step needs to return particle list. dynRelaxation Evaluate() will take particle list with updated positions and then map that back to ref points to update their position. Curve-by-points will update automatically, no need to make spring length back to CBPs. on evaluate all we are doing is updating the position of existing ref points so refresh should be fairly quick, update ref points individually or as a group. question is do we need nested transactions to do this. Need to ask Stephen/Peter about threading.
basic use cases: