Experimental Studies of Graph Traversal Algorithms

We conduct an experimental evaluation of all major online graph traversal algorithms. This includes many simple natural algorithms as well as more sophisticated strategies. The observations we made watching the animated algorithms explore the graphs in the interactive experiments motivated us to introduce some variants of the original algorithms. Since the theoretical bounds for deterministic online algorithms are rather bad and no better bounds for randomized algorithms are known, our work helps to provide a better insight into the practical performance of these algorithms on various graph families. It is to observe that all the tested algorithm have a performance very close to the optimum offline algorithm in a huge family of random graphs. Only few very specific lower bound examples cause bad results.