Model predictive control based trajectory optimization for nap-of-the-earth (NOE) flight including obstacle avoidance

This work presents a model predictive control based trajectory optimization method for nap-of-the-earth (NOE) flight including obstacle avoidance, emphasizing the mission objective of low altitude at high speed. A NOE trajectory reference is generated over a subspace of the terrain. It is then inserted into the cost function and the resulting trajectory tracking error term is weighted for more precise longitudinal tracking than lateral tracking through the introduction of the TF/TA ratio. Obstacle avoidance including preclusion of ground collision is accomplished through the establishment of hard state constraints. These state constraints create a 'safe envelope' within which the optimal trajectory can be found. Steps are taken to reduce complexity in the optimization problem including perturbational linearization in the prediction model generation and the use of control basis functions. Preliminary results over a variety of sample terrains are provided to show the mission objective of low altitude and high speed was met satisfactorily without terrain or obstacle collision, however, methods to preclude or deal with infeasibility must be investigated as speed is increased to and past 30 knots.

I. Introduction

SURVIVABILITY is a primary research objective for unmanned aerial vehicles. It is inherently obvious that as threat exposure increases, so does the probability of vehicle attrition. To effectively counter threat exposure, suitable guidance and control algorithms can take advantage of terrain masking through Nap-of-the-Earth (NOE) flight (less than 10m AGL) while simultaneously flying as fast as possible to enhance survivability if detected. This high speed requirement differentiates our research from traditional autonomous NOE flight which typically sacrifices velocity to attain fine altitude tracking. For such missions that involve cluttered, dangerous terrain, our high-speed, NOE flight objective translates to a highly constrained control problem. It is assumed that at low altitudes, the vehicle will be operating with an incomplete obstacle map which will be continuously updated with real-time sensor data. This motivates the need for an efficient algorithm for introducing new obstacle knowledge into the trajectory optimization to allow dynamic trajectory replanning.