### Summary

This paper discusses a hybrid system for handling collisions. The author asserts that impulse-based simulation and constraint-based simulation solve orthogonal problems. As such, an optimal system would use both methods. For parts of the simulation that deal with many collisions, impulse based methods are superior as collisions are dealt with in a more robust manner. Constraint-based methods should be used in situations such as a hinge joint where impulse methods are extraordinarily slow as many collisions need to be dealt with. Constraint-based methods, however, are very simple in these cases.

One difficulty in such a hybrid system is determining when to switch between paradigms. As an example, when a book is set on a table, an impulse based system can be used to realistically simulate the book settling in to a rest state. From that point onward a constraint based system can be used to more efficiently keep the book resting on the table. Unfortunately, the authors offer no suggestions on how to actually solve this problem.

Another problem is collision detection. In impulse based systems, ballistic trajectories are used to simplify the collision detection process. In constraint based systems objects may follow other trajectories, complicating collision detection. Once again, unfortunately, the authors offer no possible solutions to this problem.