Scenario 4. Several landmarks and obstacles

    

Description: In this scenario there are also a few non visible long obstacles between some landmarks that completely block the shortest path from the starting point to the target landmark.

Task: Reach the target landmark avoiding obstacles and building a map of the environment, and using it to compute diverting targets.

Results: the robot did successfully encode the obstacles on the topological map and used it to compute diverting targets. In 58% of the trials only one diverting target was computed in order to avoid a long obstacle blocking the path; the rest of the obstacles were avoided by the Pilot system, with no need to compute more diverting targets. In 18% of the trials, however, it was necessary to compute another diverting target, since the Pilot found the path blocked again by a long obstacle. On the other hand, in 24% of the trials, the Pilot was able to avoid the long obstacles, but did not realize that they were such long obstacles. This situation happened when the crash points with the long obstacle were not close enough to each other or to the landmarks, so they were considered as independent obstacles. Thus, when the Pilot tried to avoid these ``point obstacles'', it was actually avoiding the long obstacle, without realizing it. In such situations, the robot reached the target without having to compute any diverting target. Bids for moving actions were distributed very similarly as in the two previous scenarios. The only difference is that the Rescuer also won some bids (actually, it only wins one bid for stopping the robot each time it asks for a diverting target). Regarding bids for looking actions, now the Target Tracker also won a few bids to look towards the target to decrease its location's imprecision. Be it for these actions or because the scenario was not complex enough, the imprecision was never high enough so that the Rescuer had to bid for looking actions. Again, some of the trials failed because of the error on the target's location. In Section 6.7 we describe in detail one trial in this scenario.