Page 294 - Introduction to Autonomous Mobile Robots

Planning and Navigation

Figure 6.11 279
Shape of the bubbles around the vehicle (courtesy of Raja Chatila [85]).

vehicles and so we focus only on the bubble band extension made by Khatib, Jaouni, Cha-
tila, and Laumod [85].
A bubble is defined as the maximum local subset of the free space around a given con-
figuration of the robot that which can be traveled in any direction without collision. The
bubble is generated using a simplified model of the robot in conjunction with range infor-
mation available in the robot’s map. Even with a simplified model of the robot’s geometry,
it is possible to take into account the actual shape of the robot when calculating the bubble’s
size (figure 6.11). Given such bubbles, a band or string of bubbles can be used along the
trajectory from the robot’s initial position to its goal position to show the robot’s expected
free space throughout its path (see figure 6.12).
Clearly, computing the bubble band requires a global map and a global path planner.
Once the path planner’s initial trajectory has been computed and the bubble band is calcu-
lated, then modification of the planned trajectory ensues. The bubble band takes into
account forces from modeled objects and internal forces. These internal forces try to mini-
mize the “slack” (energy) between adjacent bubbles. This process, plus a final smoothing
operation, makes the trajectory smooth in the sense that the robot’s free space will change
as smoothly as possible during path execution.
Of course, so far this is more akin to path optimization than obstacle avoidance. The
obstacle avoidance aspect of the bubble band strategy comes into play during robot motion.
As the robot encounters unforeseen sensor values, the bubble band model is used to deflect
the robot from its originally intended path in a way that minimizes bubble band tension.
An advantage of the bubble band technique is that one can account for the actual dimen-
sions of the robot. However, the method is most applicable only when the environment con-
figuration is well-known ahead of time, just as with off-line path-planning techniques.