[wireties]

This a cool academic exercise but since FIRST is about real-world practices, I'll add in another 2 cents.

"For example, lets say that in your autonomous, you want to watch out for nearby obstacles such as robots. When the rangefinder picks them up, you could store their location relative, a time, etc. Then, as objects are no longer detected for a certain amount of time, pass out of range, etc., you delete them. Now, you could have a stack, find the first unused "object", store in it, loop through them, make sure they are within range, mark them as unused, etc."

Such a data structure would be fixed in size. In a real-time embedded environment, this would never be dynamically allocated (except maybe the entire memory area needed and then only during startup). One would set up an array of these objects and build a simple library to "allocate" them and "return" them to the pool. This would be a constant order operation. Non-deterministic functions (like malloc) are not a good idea in time sensitive software/firmware. Random-sized dynamic allocations can't be done deterministically (easily) because any algorithm to traverse the free-list is O(f(n)) and must be protected using mutual exclusion. This is the case in every OS I've ever used (and that is a long list).

HTH