Experimental investigation of voting schemes for fusion of redundant purposive modules

摘要

. For efficient implementation, specialized purposive modules incorporate very specific assumptions based on knowledge about the given task and environment. This makes them vulnerable to deviations that cause these assumption to become invalid. Our main objective is to construct modules that are less vulnerable to invalidation of such specific assumptions. In this paper we investigate how voting algorithms can be used to improve the reliability of purposive modules. Our is similar to N-version programming, where a number of functionally redundant modules vote for a set of possible actions. An appropriate action is then selected based on the received votes. In order to investigate our conjecture of improved reliability using voting schemes, two sets of experiments were conducted. In the first, a team of obstacle avoidance modules vote to guide a mobile robot platform in the most appropriate direction. In the second, four object tracking algorithms are integrated using a voting scheme to...