Abstract: | This dissertation describes a real-world implementation of the action selection mechanism developed by Rosenblatt and Payton at the Hughes Artificial Intelligence Center. The mechanism can be viewed as a fine-grained approach to Brooks' subsumption architecture, whereby task-achieving behaviours are broken into many smaller decision-making units. Computer programs for an autonomous Lego vehicle were developed and used to give the vehicle the ability to survive and achieve simple goals within a laboratory environment. Comparison is made with the results of an implementation of Pattie Maes' Spreading Activation Network.
|