``Wreckless'' driving

Figure 1.13: Reckless driving: (a) Using a planning algorithm to drive a car quickly through an obstacle course [199]. (b) A contender developed by the Red Team from Carnegie Mellon University in the DARPA Grand Challenge for autonomous vehicles driving at high speeds over rugged terrain (courtesy of the Red Team).

Now consider driving the car at high speeds. As the speed increases, the car must be treated as a dynamical system due to momentum. The car is no longer able to instantaneously start and stop, which was reasonable for parking problems. Although there exist planning algorithms that address such issues, there are still many unsolved research problems. The impact on industry has not yet reached the level achieved by ordinary motion planning, as shown in Figures 1.3 and 1.4. By considering dynamics in the design process, performance and safety evaluations can be performed before constructing the vehicle. Figure 1.13 shows a solution computed by a planning algorithm that determines how to steer a car at high speeds through a town while avoiding collisions with buildings. A planning algorithm could even be used to assess whether a sports utility vehicle tumbles sideways when stopping too quickly. Tremendous time and costs can be spared by determining design flaws early in the development process via simulations and planning. One related problem is verification, in which a mechanical system design must be thoroughly tested to make sure that it performs as expected in spite of all possible problems that could go wrong during its use. Planning algorithms can also help in this process. For example, the algorithm can try to violently crash a vehicle, thereby establishing that a better design is needed.

Aside from aiding in the design process, planning algorithms that consider dynamics can be directly embedded into robotic systems. Figure 1.13b shows an application that involves a difficult combination of most of the issues mentioned so far. Driving across rugged, unknown terrain at high speeds involves dynamics, uncertainties, and obstacle avoidance. Numerous unsolved research problems remain in this context.