12.1 General Methods

This section presents planning methods for the problems introduced in Section 11.1. They are based mainly on general-purpose dynamic programming, without exploiting any particular structure to the problem. Therefore, their application is limited to small state spaces; nevertheless, they are worth covering because of their extreme generality. The basic idea is to use either the nondeterministic or probabilistic I-map to express the problem entirely in terms of the derived I-space, ${\cal I}_{ndet}$ or ${\cal I}_{prob}$, respectively. Once the derived information transition equation (recall Section 11.2.1) is defined, it can be imagined that ${\cal I}_{ndet}$ or ${\cal I}_{prob}$ is a state space in which perfect state measurements are obtained during execution (because the I-state is always known).