Page 385 - Introduction to AI Robotics

368

a. 10 Metric Path Planning

b.

Figure 10.12 a.) Trulla output and photo of Clementine navigating around cord. b.)
Trulla output and photo of Clementine navigating over the cord.

Now it has to turn and move again. A resulting see-sawing set of motions
emerges that can go on for minutes. This wastes time and energy, as well as
makes the robot appear to be unintelligent. The problem is exacerbated by
non-holonomic vehicles which may have to back up in order to turn to reach
a particular location. Backing up almost always introduces more errors in
navigation.
To handle subgoal obsession, many roboticists program into their move-to-
goal behaviors a tolerance on the terminating condition for reaching a goal.
A common heuristic for holonomic robots is to place a tolerance of +/- the
width of the robot. So if a cylindrical, holonomic robot with a diameter of 1
meter is given a goal of (35, 50), it will stop when 34.5 < < 35.5 and 49.5 y