Multiresolution and Adaptive Path Planning for Maneuver of Micro-Air-Vehicles in Urban Environments

Abstract

This paper addresses adaptive path planning and control of autonomous micro-air-vehicles (MAVs) in urban environments. The objective is to navigate a MAV from one point to another in an unknown urban environment. A key consideration in such environments is the ability to estimate the locations of obstacles that may lie in the path of the vehicle. In this study, obstacle estimation is achieved using an adaptive multiresolution-based learning algorithm. This algorithm generates an approximation of the environment based on estimates of the locations of points on the surface of an obstacle, typically obtained using structure from motion analyses. The overall flight control problem is formulated as a receding horizon control problem, which entails extremizing a performance functional over a moving window in time to determine the optimal control inputs. Receding horizon control represents an extremely general framework for addressing the problem of autonomous flight in urban environments. Most importantly, it allows for the incorporation of multiple constraints in the optimization problem so that obstacle avoidance can be achieved by enforcing constraints on the aircraft states. Receding horizon control is achieved by solving many small-scale optimization problems. Considerable computational savings can be realized compared to solving for optimal control strategies over the full time scale of interest. In addition, it is possible to update the constraints based on new information such as updated estimates of obstacle shape. The main objective of this paper is to study the receding horizon control problem in which the adaptive learning algorithm is used to estimate the obstacle map from visual data. An important consideration in future studies will be the stability of this control algorithm as it is well understood that receding horizon control approaches do not enjoy the same stability properties as infinite-horizon optimal control methods. Some specific path planning and control strategies that fall within the receding horizon control framework are discussed, and examples of terrain estimation using the adaptive learning algorithm are presented.