Organizational Unit:
Daniel Guggenheim School of Aerospace Engineering
Daniel Guggenheim School of Aerospace Engineering
Permanent Link
Research Organization Registry ID
Description
Previous Names
Parent Organization
Parent Organization
Organizational Unit
Includes Organization(s)
Organizational Unit
Organizational Unit
Organizational Unit
Organizational Unit
Organizational Unit
ArchiveSpace Name Record
Publication Search Results
Now showing
1 - 6 of 6
-
ItemConcurrent learning for convergence in adaptive control without persistency of excitation(Georgia Institute of Technology, 2010-11-11) Chowdhary, GirishModel Reference Adaptive Control (MRAC) is a widely studied adaptive control methodology that aims to ensure that a nonlinear plant with significant modeling uncertainty behaves like a chosen reference model. MRAC methods attempt to achieve this by representing the modeling uncertainty as a weighted combination of known nonlinear functions, and using a weight update law that ensures weights take on values such that the effect of the uncertainty is mitigated. If the adaptive weights do arrive at an ideal value that best represent the uncertainty, significant performance and robustness gains can be realized. However, most MRAC adaptive laws use only instantaneous data for adaptation and can only guarantee that the weights arrive at these ideal values if and only if the plant states are Persistently Exciting (PE). The condition on PE reference input is restrictive and often infeasible to implement or monitor online. Consequently, parameter convergence cannot be guaranteed in practice for many adaptive control applications. Hence it is often observed that traditional adaptive controllers do not exhibit long-term-learning and global uncertainty parametrization. That is, they exhibit little performance gain even when the system tracks a repeated command. This thesis presents a novel approach to adaptive control that relies on using current and recorded data concurrently for adaptation. The thesis shows that for a concurrent learning adaptive controller, a verifiable condition on the linear independence of the recorded data is sufficient to guarantee that weights arrive at their ideal values even when the system states are not PE. The thesis also shows that the same condition can guarantee exponential tracking error and weight error convergence to zero, thereby allowing the adaptive controller to recover the desired transient response and robustness properties of the chosen reference models and to exhibit long-term-learning. This condition is found to be less restrictive and easier to verify online than the condition on persistently exciting exogenous input required by traditional adaptive laws that use only instantaneous data for adaptation. The concept is explored for several adaptive control architectures, including neuro-adaptive flight control, where a neural network is used as the adaptive element. The performance gains are justified theoretically using Lyapunov based arguments, and demonstrated experimentally through flight-testing on Unmanned Aerial Systems.
-
ItemFlight Test Validation of a Neural Network based Long Term Learning Adaptive Flight Controller(Georgia Institute of Technology, 2009-08) Chowdhary, Girish ; Johnson, Eric N.The purpose of this paper is to present and analyze flight test results of a Long Term Learning Adaptive Flight Controller implemented on a rotorcraft and a fixed wing Unmanned Aerial Vehicle. The adaptive control architecture used is based on a proven Model Reference Adaptive Control (MRAC) architecture employing a Neural Network as the adaptive element. The method employed for training the Neural Network for these flight tests is unique since it uses current (online) as well as stored (background) information concurrently for adaptation. This ability allows the adaptive element to simulate long term memory by retaining specifically stored input output data pairs and using them for concurrent adaptation. Furthermore, the structure of the adaptive law ensures that concurrent training on past data does not affect the responsiveness of the adaptive element to current data. The results show that the concurrent use of current and background data does not affect the practical stability properties of the MRAC control architecture. The results also confirm expected improvements in performance.
-
ItemReal-Time System Identification of a Small Multi-Engine Aircraft(Georgia Institute of Technology, 2009-08) DeBusk, Wesley M. ; Chowdhary, Girish ; Johnson, Eric N.In-flight identification of an aircraft's dynamic model can benefit adaptive control schemes by providing estimates of aerodynamic stability derivatives in real time. This information is useful when the dynamic model changes severely in flight such as when faults and failures occur. Moreover a continuously updating model of the aircraft dynamics can be used to monitor the performance of onboard controllers. Flight test data was collected using a sum of sines input implemented in closed loop on a twin engine, fixed wing, Unmanned Aerial Vehicle. This data has been used to estimate a complete six degree of freedom aircraft linear model using the recursive Fourier Transform Regression method in frequency domain. The methods presented in this paper have been successfully validated using computer simulation and real flight data. This paper shows the feasibility of using the frequency domain Fourier Transform Regression method for real time parameter identification.
-
ItemIndoor Navigation for Unmanned Aerial Vehicles(Georgia Institute of Technology, 2009-08) Sobers, D. Michael Jr. ; Chowdhary, Girish ; Johnson, Eric N.The ability for vehicles to navigate unknown environments is critical for autonomous operation. Mapping of a vehicle's environment and self-localization within that environ- ment are especially difficult for an Unmanned Aerial Vehicle (UAV) due to the complexity of UAV attitude and motion dynamics, as well as interference from external influences such as wind. By using a stable vehicle platform and taking advantage of the geometric structure typical of most indoor environments, the complexity of the localization and mapping problem can be reduced. Interior wall and obstacle location can be measured using low-cost range sensors. Relative vehicle location within the mapped environment can then be determined. By alternating between mapping and localization, a vehicle can explore its environment autonomously. This paper examines available low-cost range sensors for suitability in solving the mapping and localization problem. A control system and navigation algorithm are developed to perform mapping of indoor environments and localization. Simulation and experimental results are provided to determine feasibility of the proposed approach to indoor navigation.
-
ItemGeorgia Tech Aerial Robotics Team: 2009 International Aerial Robotics Competition Entry(Georgia Institute of Technology, 2009-07) Chowdhary, Girish ; Christmann, Hans Claus ; Johnson, Eric N. ; Salaün, Erwan ; Sobers, D. Michael Jr.This paper examines the use of low-cost range and target identification sensors on a stable flying vehicle for suitability in solving the 5th Mission proposed for the 2009 International Aerial Robotics Competition. The ability for vehicles to navigate unknown environments is critical for autonomous operation. Mapping of a vehicle's environment and self-localization within that environment are especially difficult for an Unmanned Aerial Vehicle (UAV) due to the complexity of UAV attitude and motion dynamics. Using a stable vehicle platform and taking advantage of the geometric structure typical of most indoor environments reduces the complexity of the localization and mapping problem to the point that wall and obstacle location can be determined using low-cost range sensors. Target identification is accomplished remotely using an onboard video camera with a radio transmitter. Thus complex and time-consuming image processing routines are run on a more powerful computer, enabling further miniaturization of the flight vehicle.
-
ItemAdaptive Neural Network Flight Control Using both Current and Recorded Data(Georgia Institute of Technology, 2007-08) Chowdhary, Girish ; Johnson, Eric N.Modern aerospace vehicles are expected to perform beyond their conventional flight envelopes and exhibit the robustness and adaptability to operate in uncertain environments. Augmenting proven lower level control algorithms with adaptive elements that exhibit long term learning could help in achieving better adaptation performance while performing aggressive maneuvers. The current adaptive methodologies which use Neural Network based control methods use only the instantaneous states to tune the adaptive gains. This results in a rank one limitation on the adaptive law. In this paper we propose a novel approach to adaptive control, which uses the current or the online information as well as stored or background information for adaptation. We show that using a combined online and background learning approach it is possible to overcome the rank one limitation on the adaptive law resulting in faster adaptation to the unknown dynamics. Furthermore, we show that using combined online and background learning methods it is possible to guarantee long term learning in the adaptive flight controller, which enhances performance of the controller when it encounters a maneuver that has been performed in the past. We use Lyapunov based methods for showing boundedness of all signals for a proposed method. The performance of the proposed method is evaluated in the high fidelity simulation environment for the GTMAX UAS maintained by the Georgia Tech UAV lab. The simulation results show that the proposed method exhibits long term learning and faster adaptation leading to better performance of the UAS flight controller.