Page 319 - Introduction to Autonomous Mobile Robots

Chapter 6
304
and the planner. Planning is simply one small part of the executive’s nominal cycle of activ-
ities.
The idea of speeding up planning to the point that its execution time is no longer signif-
icant may seem wishful. However, using specific algorithms in particular environments,
such reductions in the cost of planning have been demonstrated. Consider the work of
Stentz [139]. Stentz designed a mobile robot control architecture for a large off-road vehi-
cle traveling over partially known terrain at high speeds. Using advanced caching tech-
niques from computer science, Stentz optimized a grassfire path-planning algorithm called
D* so that global path planning would be possible within the basic control loop of the exec-
utive.
The result, depicted in figure 6.26, is an architecture in which the local and global rep-
resentations are the same, and in which the executive has all global planning functionality
required for the problem built in. The advantage of this approach is that the robot’s actions
at every cycle are guided by a global path planner, and are therefore optimal in view of all
of the information the robot has gathered. Of course, the method is computationally chal-
lenging and will not be practical in more complex environments until processor speeds
increase even further. It also has basic limits of applicability as the size of the environment
increases, but this has not yet been a barrier when applying this method to real-world sce-
nario sizes.
The somewhat recent success of an integrated planning and execution method, D*,
underlines the fact that the designer of a robot navigation architecture must consider not
only all aspects of the robot and its environmental task but must also consider the state of
processor and memory technology. We expect that mobile robot architecture design is sure
to remain an active area of innovation for years to come. All forms of technological
progress, from robot sensor inventions to microprocessor speed increases, are likely to cat-
alyze new revolutions in mobile robot architecture as previously unimaginable tactics
become realities.