Modeling We wrote a balloon simulation without leaving the ground. It sure looks real to us. lift A real ballooner contributed a textbook model of balloon life depending on many factors of balloon state and atmosphere state. Unfortunately, we didn't learn enough about that model to attempt to implement it. control inputs Having said that we've removed the complexity of simulation of lift to the modeling of the effect of control inputs on the balloon's net energy, we didn't get very far at modelling much of that complexity. Software The Sense8 WorldUp software was mighty frustrating to work with. Aparently incompatible WorldUp versions Different versions of WorldUp are running on the Media Union machines and the VRL machines. An initial flight model developed on the relatively available MU machines caused the VRL version to first complain that the universe was generated by software of a different version, then dump core on attempting to render. Since we intend for the model to operate with the VRL BOOM, we confined further development to the VRL. Access to the VRL system was limited. WorldUp instability WorldUp frequently dumped core or locked up during working sessions. We discovered that some lockups were due to attempts to print error messages to the terminal, which STOPped the application. After discovering this, we worked around it by running WorldUp in the foreground. The frequent crash-and-dump-core events were not so easy to work around. The application mysteriously alternated between running through many minutes of development and testing, and crashing on the first attempt to do perform some flight action or another. Typically, it crashed on an input event such as a mouseclick or keypress. We might have kept meticulous notes on crash situations, but didn't. On a number of occasions, a universe that worked at the end of one day would crash out at the beginning of the next day. Collision detection We expected collision detection to be all done for us in WorldUp. To the best of our ability to determine, collision detection as such is done between bounding boxes of complex geometries or groups of geometries. First we had to figure this out, then we had to figure out a method for collision detection.

Return to main home page http://www-VRL.umich.edu/~saha/eng477/ Last update: October 14, 1996 by DS