Analysing solution quality of anytime Bag of Words object classification for a service robot

Much work can be found in the literature that theoretically analyse anytime algorithms for resource aware systems. However, only a few of them focus on practical real world applications e.g. for mobile service robots with limited resources. This paper revises some theoretical assumptions according to the needs in a tangible object classification scenario and shows an application-related mechanism that facilitates a service robot to not only be aware of but also to actively manage its resources. For this purpose, an object classification example based on Bag-of-Words (resp. Bag-of-Features) is presented. First, the underlying algorithm is modified so as to setup its execution time limit. Thereto, the algorithm is combined with a task-pair. Second, the performance profile of the algorithm is determined by analysing the influence of execution time limits on resulting solution quality.