A study of multiple behavior implementations in connection with the utility manifold method for behavioral organization

Journal article, 2004

In this paper, the performance of the utility manifold (UM) method for behavioral organization is investigated. The method is applied to a case involving strongly non-trivial selection between four different behaviors, in order to generate an overall task of navigation for a simulated wheeled robot. The results of the investigation show that the UM method was easily able to achieve the overall navigation task, by generating appropriate selection between the four constituent behaviors.
A desirable property of any method for behavioral organization is the ability to organize different behaviors regardless of their specific implementation. This property is investigated for the UM method, by testing it against two different versions for each of the four constituent behaviors, i.e. a total of 16 different combinations.