Next: Navigation functions Up: 8.2 Discrete State Spaces Previous: Feasibility and optimality
It conveniently turns out that tools for computing a feedback plan were already given in Chapter 2. Methods such as Dijkstra's algorithm and value iteration produce information as a side effect that can be used to represent a feedback plan. This section explains how this information is converted into a feedback plan. To achieve this, a feedback plan will be alternatively expressed as a potential function over the state space (recall potential functions from Section 5.4.3). The potential values are computed by planning algorithms and can be used to recover the appropriate actions during execution. In some cases, an optimal feedback plan can even be represented using potential functions.