| Summary: | Thesis (S.M.)--Massachusetts Institute of Technology, Dept. of Aeronautics and Astronautics, 2012. === Cataloged from PDF version of thesis. === Includes bibliographical references (p. 107-111). === This thesis is motivated by the problem of fixed-wing flight through obstacles using only on-board sensing. To that end, we propose novel algorithms in trajectory generation for fixed-wing vehicles, state estimation in unstructured 3D environments, and planning under uncertainty. Aggressive flight through obstacles using on-board sensing involves nontrivial dynamics, spatially varying measurement properties, and obstacle constraints. To make the planning problem tractable, we restrict the motion plan to a nominal trajectory stabilized with an approximately linear estimator and controller. This restriction allows us to predict distributions over future states given a candidate nominal trajectory. Using these distributions to ensure a bounded probability of collision, the algorithm incrementally constructs a graph of trajectories through state space, while efficiently searching over candidate paths through the graph at each iteration. This process results in a search tree in belief space that provably converges to the optimal path. We analyze the algorithm theoretically and also provide simulation results demonstrating its utility for balancing information gathering to reduce uncertainty and finding low cost paths. Our state estimation method is driven by an inertial measurement unit (IMU) and a planar laser range finder and is suitable for use in real-time on a fixed-wing micro air vehicle (MAV). The algorithm is capable of maintaining accurate state estimates during aggressive flight in unstructured 3D environments without the use of an external positioning system. The localization algorithm is based on an extension of the Gaussian Particle Filter. We partition the state according to measurement independence relationships and then calculate a pseudo-linear update which allows us to use 25x fewer particles than a naive implementation to achieve similar accuracy in the state estimate. Using a multi-step forward fitting method we are able to identify the noise parameters of the IMU leading to high quality predictions of the uncertainty associated with the process model. Our process and measurement models integrate naturally with an exponential coordinates representation of the attitude uncertainty. We demonstrate our algorithms experimentally on a fixed-wing vehicle flying in a challenging indoor environment. The algorithm for generating the trajectories used in the planning process computes a transverse polynomial offset from a nominal Dubins path. The polynomial offset allows us to explicitly specify transverse derivatives in terms of linear equality constraints on the coefficients of the polynomial, and minimize transverse derivatives by using a Quadratic Program (QP) on the polynomial coefficients. This results in a computationally cheap method for generating paths with continuous heading, roll angle, and roll rate for the fixed-wing vehicle, which is fast enough to run in the inner loop of the RRBT. === by Adam Parker Bry. === S.M.
|
|---|
