Pilot System

The Pilot is able to safely command the motors that control the robot to move in a given direction. It bids for motion control to avoid obstacles, and also for the control of the camera to look forward in order to detect possible obstacles. Although this system is not the focus of this thesis, we have had to develop a simple Pilot in order to test our Navigation system.

For obstacle avoidance, it uses the information coming from the Vision system and the information stored in the Visual Memory (described in the next section), applying an obstacle growing technique. The obstacles are grown a given size to define forbidden areas occupied by the obstacles. The obstacles are represented as points (for landmarks and simple obstacles) and lines (for linear obstacles between landmarks), which, after growing them, become circles and rounded rectangles, respectively. In our case, the growing size is the diameter of the robot. An example of how the obstacles are grown is shown in Figure 4.3. The Pilot uses a simple obstacle avoidance algorithm. It checks whether the robot is about to enter any of the forbidden areas associated to the obstacles. If the robot is in such a situation, the Pilot bids to modify the trajectory in order to avoid the obstacle. The modified trajectory is tangential to the grown obstacle to be avoided. Since obstacle avoidance is of maximal importance, the bid should be higher than the other systems. However, it should not be set to the highest possible value, 1, so that there is the possibility of adding a new system that overrides the Pilot (e.g. a teleoperation system). If the robot is in a safe area, the Pilot does not bid at all.

Figure 4.3: Growing obstacles. Points and solid lines are the obstacles; dotted lines show grown obstacles

Regarding the bids for camera control, it is based on a function that increases the bid depending on the distance traveled since the last time the robot looked forward:

$$bid(look(ahead))=\left( \frac{dist\_since\_last\_look}{max\_dist\_not\_looking} \right)^{exp}$$ (5)

where $max\_dist\_not\_looking$ is the maximum distance allowed to travel without looking ahead, and $exp$ defines the increasing shape of the bidding function.

The Pilot also informs the Navigation system and the Visual Memory about any obstacle it detects. Whenever it detects a single obstacle (i.e. it bumps into it), it stores the obstacle's location in the Visual Memory, and checks whether it can be part of a larger linear obstacle. Such linear obstacles are detected when a series of single obstacles have been detected along the line connecting two landmarks and the distance between these obstacles is below a given threshold. If this is the case, the Pilot informs the Navigation system about the presence of a blocking obstacle between two landmarks.