Requires Windows, Mac, or Linux
Got an idea!This require I think from tens to thousends MB (so even GB) of RAM memory (I don't believe that the normal one is suitable):Start the simulation: nothing appear. Data are stored. After some time (set by the user), maybe 10 years, maybe 1 year (obviously of the simulation), maybe dipending on the time step, the stored data are converted in images. The result? A simulation that runs with data of the future really accurate. The problem is the performance that this buffering sucks: data must be stored, which can take A LOT of RAM, especially in big systems, and then reloaded and calculated (the coordinates of every object must be stored frame by frame. For a ten-object simulation, which is really small, there are 30 numbers with lots of decimal digits every frame). I think that it could be optimized for dual/quad cores (with a dedicated core the RAM occupied will be smaller).Moreover if you sacrifice your knowledge of the future you can make the simulation with auto time step more "linear" using the buffer.Pros: A future more accurate - Projected paths way more accurate - Collisions predictable for sure within some seconds/minutes (according to your CPU/will) - Auto Time step more fluid (scarifying your knowledge of the future).Cons: Performance needs increasing exponentially with the number of objects.EDIT: It also permit to move the simulation forward or backward maybe with a time bar. E.g:-3 -2 -1 0 1 2 3|----------|----------|----------|-O-------|----------|----------|