The 21st Century COE Program
Center of Excellence for Research and Education on Complex Functional Mechanical Systems

---

トップページ

本拠点について

研究会等のお知らせ

募集

資料

拠点内専用

English

制御設計G特別研究会

開催概要

プログラム

講演要旨

Stochastic Optimized Monocular Vision-Based Guidance Design

Yoko Watanabe (Graduate Research Assistant, Georgia Institute of Technology)

This presentation discusses a vehicle guidance design to achieve given missions using a single passive 2-D vision sensor. An extended Kalman filter (EKF) is applied for relative state estimation, and the estimate is fed back to the guidance system. It is very common to derive a guidance law by optimizing an appropriate cost assuming full-state information. However, for vision-based control, it is well-known that estimation performance significantly depends on sensor motion relative to a target. Therefore, we suggest a vision-based guidance design which optimizes an expectation value of the cost under a condition of using the EKF-based navigation. By this design, sensor trajectory optimization which improves estimation accuracy is included as a part of vehicle guidance design. However, the resulting optimization problem becomes nonlinear due to the vision-based measurement, and its optimal solution can only be obtained numerically. For real-time implementation of the guidance system, a sub-optimal solution is derived by using one-step-ahead optimization method. The suggested guidance design is applied to vision-based target tracking problem, and simulation results verify that it improves the estimation accuracy and hence improves overall guidance performance.

Orbit control of the Hill's Equation with Impulse Inputs

Akihiro Ichimura (Dept. of Aeronautics and Astronautics, Graduate School of Engineering, Kyoto University)

In this talk orbital rendezvous of a chaser with a target spacecraft in a circular orbit is considered. The Hill's equation describes the realtive motion of the chaser with respect to the target and possess periodic solutions. They are useful for flyaround and rendezvous-docking missions. We consider the minimum-fuel orbit transfer problem from a periodic orbit to another by impulsive velocity changes. We derive optimal three-impulse controllers and then propose feedback controllers which have performances close to the optimal controllers. Simulation results will be given to illustrate our analysis and design procedure.

Adaptive Control and Vision-based Control Research Highlights

Eric N. Johnson (Lockheed Martin Assistant Professor of Avionics Integration Georgia Institute of Technology)

Today, autonomous systems are far less reliable or capable than their manned counterparts. This talk covers two approaches to "get back" some of the capability lost when attempting to operate an aircraft without a pilot on board. First, the ability of a human pilot to adapt to changes in the flight characteristics of an aircraft is addressed with advances in adaptive flight control methods. Specifically, online trained artificial neural networks are discussed. Here, a dynamic inversion control law is augmented with a neural network to correct for model errors in each degree of freedom. Additionally, actuator nonlinearities are accounted for in the theory. Full-envelope flight test results on helicopters, ducted-fans, and aerobatic airplanes will be described. Second, the utilization of vision sensors for guidance, navigation, and control is explored. This will include descriptions of theory and algorithms, as will flight test and demonstration of autonomous aircraft with these new capabilities. Theory includes advances in image processing, estimation theory, and guidance policies. The demonstrated methods include: vision-aided inertial navigation, vision-based air-to-air tracking methods, and vision-only guidance, navigation, and control.