Page 152 - Innovations in Intelligent Machines

Evolution-based Dynamic Path Planning for Autonomous Vehicles 143
160
Expected value of loss function 100
140
120

80
60

40
20
0
0 5 10 15 20 25 30 35 40
Time sample

Fig. 25. Evolution of dynamic path planning with moving obstacles

We present an evolutionary algorithm suitable for dynamic path plan-
ning problems. The algorithm was developed as part of the Evolution-based
Cooperative Planning System for teams of autonomous vehicles. The algo-
rithm is used to find an optimal path that maximizes an objective function.
This function is formulated using the stochastic world model to capture the
dynamics and uncertainties in the system. The algorithm has been applied to
path planning for UAVs in real wind fields and predicted icing conditions [22].
Extensions to apply this algorithm to the coupled task of the path planning
problem for multiple cooperating vehicles are reported in [19].
Simulation results demonstrate that the path planning algorithm can
provide computationally feasible effective solutions to all of the path plan-
ning problems which include planning with timing constraints and dynamic
planning with moving targets and obstacles. Even though there are some
uncertainties in the knowledge of the environment, the algorithm can generate
feasible paths which are within the capabilities of the vehicle to complete all
tasks and to avoid collision with the obstacles. During the mission, the planner
is able to quickly adapt the path in response to changes in the environment.

8 Acknowledgments

The research presented in this paper is partially funded by the Washington
Technology Center. The simulation software is provided by the Boeing
Company. Professor Emeritus Juris Vagners at the University of Washington
provided direction and advice for this research.