To add capability to unmanned ground vehicle systems the GPS and Vehicle Dynamics Lab at Auburn University is developing new path planning algorithms. These algorithms allow autonomous vehicles to navigate more complicated obstacle fields or allow the driver of tele-operated vehicles to assign overall goals for fleets of vehicles instead of remotely controlling each vehicle individually. The algorithms operate by using artificial potential fields to mimic how water flows through an obstacle field.
For many years, potential fields have been used to model fluid flow. These fields do not suffer from the local minima of the more general potential fields and also are often smoother than traditional potential fields. Another advantage fluid potential fields have is the ability to model many complex shapes with a wide range of primitive objects. Figure 1 shows fours of these building block flows: clockwise – uniform flow implying straight driving, sink (or source) flow for goals or starting locations, flow around a cylinder for obstacle avoidance, and flow in a sector for turning corners. Figure 2 shows multiple building block flows assembled into a complex world map. The starting location is (10,0). There are two goals (0,10) and (0,0); (0,10) is designated as more important, so most flow lines lead to it. The vehicles follow these lines to reach their goal.
Real-world scenario with potential solution
Potential field building blocks