Navigation Tasks

For a given environment we consider two different navigation tasks. Each one of them with a different level of complexity. The best parameter set may change depending on the complexity of the task. We conjecture that the parameters found depend mainly on the complexity of the navigation task and not so much on the structure of the overall environment. This complexity is dependent, though not equal, to the cartographic complexity of the world in which the agent moves, and is based on the following factors:

1. Number of visible landmarks at any time
2. Density of obstacles in the region of navigation
3. Visibility of the target at any time

Using this notion of navigational complexity, the total space of all navigation tasks can be split into two representative classes: going towards the target free of obstacles, and reaching targets located behind obstacles. In our experiments we use clusters $C_{1}$ (encircled targets in Figure 5.7) and $C_{2}$ (encircled targets in Figure 5.8) as representatives of the two task complexity classes. The best parameter set is determined for both these classes. The aim of the experiments is to endow the Navigation system of the robot with the capability to switch between these two parameter sets according to the actual task complexity it is facing.

Figure 5.7: Cluster C1

Figure 5.8: Cluster C2