Approximate Dynamic Programming for Dynamic Vehicle Routing by Marlin Wolf Ulmer

Author:Marlin Wolf Ulmer
Language: eng
Format: epub
Publisher: Springer International Publishing, Cham

(6.14)

The weighting favors LTs with large intervals in the beginning for a fast first approximation. Later, LTs with small intervals are weighted higher to achieve a more differentiated evaluation. The weight for a LT is increased by a relatively small variance and bias and a large number of observations. In the beginning, the frequently visited entries in the LTs with large intervals allow a fast first estimation of the value function. During the subsequent approximation process, the weights of the more detailed LTs with small intervals increase because of the relatively small variance and bias. Further, WLT allows avoiding ineffective LT-areas. For instance, areas in the LT providing relatively low expected future rewards are early excluded from the approximation process. Hence, the approximation is focused on the effective areas. WLT may allow a faster approximation in the beginning and high quality solutions in the end without any tuning necessary. Nevertheless, WLT leads to increased memory consumption due to the large number of entries. Instead of a single LT, L LTs of different partitioning levels are required.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.