Mutual Information Theory And Applications In The Process Of Robot Active Olfaction

Robot active olfaction of complex and dangerous environment has important academic value and application prospects in industrial production, social security, military affairs and so on. Robot odor source search is an active olfaction. Many scholars at home and abroad have done research on the algorithm of robot active olfaction. With the development of society and technology and the increase of robot researchers( providers) and robot types, various and specific robots' search strategies or methods will be inevitably more and more. It is a tend to do research on the evaluation methods of robot search strategy.The main contents of this paper and the results are as follows :(1) The Brown entropy is more accurate and reliable than the discrete entropy calculated by the sample variance from its definition, if the sample size is small. The confidence intervals of Brown entropies are narrower than those of the sample discrete entropies, where the sample size is small. Our new empirical result is that “the minimum error entropy(MEE) criterion can outperform the traditional mean square error criterion” reported by many recent studies is not always valid.(2) The Brown entropy can be revised to its “theoretical discrete entropy” calculated by the population parameter. The revised expressions are helpful especially for a small sample size.(3) This paper proposes a way to evaluate odor source search strategy of multi-robot collaborative work by using multilevel fuzzy integrated evaluation method of fuzzy mathematics. Evaluate the performance of different multi-robot odor source search algorithms by analyzing the discrete entropy and mutual information between each two trajectories of multi-robot.