Aside from the first scenario in which the shooter's behavior remained fixed, none of the scenarios described in this section can be learned with the supervised learning technique used to learn the shooting template in Section (NNs). Instead, since the agents' behaviors must continually adapt, we need a more incremental type of learning approach. One promising option is to have several trained behaviors with an adaptive decision function wrapped around them to decide which one to use.
For example, there could be a trained template that the shooter learned based on examples from one situation and another based on other examples. Then the shooter's decision function of which template to use in a given situation could be adaptive. In order to make this a truly adaptive method, additional behaviors need to be trained periodically on the newly experienced examples.