Subhabrata Ganguli

Piloted Simulation of Fault, Detection, Isolation, and Reconfiguration Algorithms for a Civil Transport Aircraft

Honeywell Labs has been researching and developing together with NASA Langley Research Center (LaRC) algorithms for aircraft failure management and recovery. The algorithms have been integrated into a Control Upset Prevention and Recovery System (CUPRSys) that provides control law reconflguration, fault detection, fault isolation and pilot cueing. This paper describes the capabilities of CUPRSys and the results from an evaluation of CUPRSys by an experimental test pilot in the Integration Flight Deck (IFD) at LaRC. Also included in the paper are details about the IFD and the MATLAB simulation environment used for design. The piloted evaluation was performed at three ∞ight conditions and the results for one representative maneuver are presented. Pilot ratings were obtained for maneuvers for the un-failed aircraft, for the failed aircraft without reconflguration and for the failed aircraft with reconflguration. Only reduction of surface efiectiveness faults were considered. The piloted simulation results suggest that CUPRSys provides a robust control law with promising fault detection, isolation and reconflguration capability.

Aircraft Fault Detection, Isolation and Reconflguration in the Presence of Measurement Errors

This paper discusses the efiect of measurement errors on the fault detection, fault isolation and control law reconflguration algorithms that Honeywell has been researching and developing together with NASA Langley Research Center (LaRC) under NASA’s Aviation Safety and Security Program. In our previous papers, we developed fault detection, fault isolation, pilot cueing and control law reconflguration algorithms for a civil transport aircraft, and evaluated the performance of our algorithms in piloted simulation in the Integration Flight Deck facility at NASA LaRC. However, we did not explicitly evaluate the efiect of measurement errors (that is, sensor noise, bias, and dynamics) on the performance of our algorithms. In this paper, we add sensor models to LaRC’s simulation model and evaluate the performance of our algorithms in the presence of measurement errors. Each algorithm is analyzed separately, and is enhanced by redesigning and/or re-tuned as necessary. The key contribution is that we provide theoretical justiflcation for the architecture of our fault detection algorithm and discuss a systematic procedure for tuning its gains, and we adapt and re-tune our fault isolation algorithm so that it can cope with measurement errors. We also provide results from batch simulations that are representative of the achieved performance in the presence of imperfect measurements.

Extremum Seeking for Model Reference Adaptive Control

Adaptive control with predictable parameter convergence has remained a challenge for several decades. In the special case of set point adaptation, extremum seeking permits predictable parameter convergence by design. This is because persistency of excitation requirements are met by the sinusoidal perturbation that is part of the basic control design. Here we present results of extremum seeking based adaptation of model reference control of a simple roll rate model of a flxed wing aircraft. We show simulation results where adaptive tracking is achieved, and the convergence of parameters conforms to predictions from the theory of extremum seeking. In the case of actuator failure, we show that the parameter convergence proofs of extremum seeking are directly applicable.

Sensor fusion for GNSS denied navigation

We present technologies that are being developed to address the need for a navigation solution in the absence of Global Navigation Satellite Systems (GNSS) measurements. The navigation system uses sensors such as vision systems, RADARS and LIDARS with feature extraction, matching and motion estimation algorithms. We present experimental results of using scale invariant feature transform, speeded up robust features, and modified Harris feature extraction algorithms and compare the performance. We also present methods to extract lines and planes that can aid in navigation. For motion estimation we present results for visual odometry as well as simultaneous localization and mapping navigation. We experimentally verify the algorithms in both a realtime Linux framework as well as offline. We also present ongoing work in vision integrated navigation in an attitude and heading reference system as well as an extended Kalman filter framework. All the methods we present in this paper are incremental navigation methods.

Statistical verification of two non-linear real-time UAV controllers

We present a versatile statistical verification methodology and we illustrate different uses of this methodology on two examples of nonlinear real-time UAV controllers. The first example applies our statistical methodology to the verification of a computation time property for a software implementation of a high-performance controller as a function of controller state variable values. The second example illustrates our statistical verification methodology applied to finding verifiably safe flight envelopes for a class of maneuvers, again as a function of controller state variable values. We compare our approach to verification with other statistical techniques used for estimating execution times and controller performance. We close with candidate topics for future work.

Region of Attraction with Performance Bounds

In this paper we introduce the notion of region of attraction with performance bounds as a useful tool for design and analysis of adaptive control systems. The benefit of using this tool is that ensuring the closed-loop dynamics for the adaptive system within this specific region can guarantee both stability and performance requirements. In this paper, we first define a region of attraction with performance bounds for a nonlinear system. We then compare it with the existing idea of a region of attraction for a nonlinear system with an illustrative example. Next, we demonstrate the usage of this new design and analysis tool with two aircraft adaptive control examples. The first example deals with the analysis of an adaptive controller for the roll-rate dynamics of an aircraft. We show that the region of attraction with performance bounds can be estimated by solving a Sum-of-Squares (SOS) relaxation of the problem which provides an efficient computation technique. The second example deals with designing an adaptive controller for the unstable short-period dynamics of an aircraft. In this example, the adaptive controller is augmented on a baseline dynamic inversion baseline control law. We demonstrate that we can use this tool to perform adaptive control design tradeoffs in an efficient manner.