Risk-Aware Path Planning for Unmanned Aerial Systems in a Spreading Wildfire

Path planning for small unmanned aerial systems (SUAS) in the presence of poorly understood and dynamic obstacles is a challenging problem: for example, flight over a spreading wildfire. Evidential information fusion is used to estimate the current wildfire state and the resulting heat aura at flight level. This approach accounts for ignorance, which is a result of conflict among sensors operating in a harsh environment and a computational forecasting agent that uses a fire evolution model of inadequate accuracy. An SUAS is employed to visit locations of high conflict to provide additional situational awareness. Flight-level heat aura is modeled as a keepout zone with probabilistic boundaries for SUAS path planning. A novel unsupervised classification algorithm is developed to identify distinct obstacle boundaries within the estimated heat aura. Path planning is posed as a chance-constrained optimal control problem, which is transcribed to a nonlinear program via pseudospectral discretization. The results show that this approach can yield a family of solutions that elicit the risk associated with each mission design, and the appropriate choice of risk can aid in the generation of "keyhole paths."