Lifecycle management for many commercial and military helicopters involves the process of regime recognition (RR), in which the maneuver history of the aircraft is classified into individual regimes over the duration of a flight. When considering verification and validation (V&V) for regime recognition codes, a key step lies in assessing the RR code's conformance to defined classification and damage prediction accuracy requirements. The confusion matrix is typically used for this purpose; however, the confusion matrix does not quantify certain critical aspects of RR code performance and oftentimes fails to provide actionable information about how a code may be improved. This paper describes a novel set of V&V diagnostic metrics designed to highlight areas of deficiencies in the algorithm and specific paths to code improvement. These metrics leverage pattern recognition techniques to align the flown and recognized maneuvers and are proposed to supplement and enhance the V&V information obtained from the confusion matrix. Several example V&V flight sequences are provided to highlight the benefits of the proposed diagnostic metrics when used in concert with other classical V&V tools.

Cooperative transportation of payloads by multiple unmanned air vehicles has received increasing interest due to unique operational advantages. These include the portability of the individual vehicles and the scalability of the lifting strategy in the presence of differing payloads. By rigidly attaching a set of unmanned air vehicles to a payload, the control effort required to transport the payload is divided across the vehicles. In the presence of uncertainty about a payload's mass and inertial characteristics, there is no inherent flightworthiness guarantee for a specific connected unmanned air vehicle configuration. This Paper describes a method for determining on-ground flightworthiness of the unmanned air vehicle-payload system while making minimal assumptions about inertial properties of the payload or the attachment configuration of the unmanned air vehicles. The probabilistic model itself is initialized and updated (built) by the algorithm. Within the model, the vehicles are positioned. Building upon prior theoretical developments, this Paper explores the differing sources of error in the estimates and experimentally validates the algorithm through a series of tests using prototype modular vehicles. Overall, simulation and test results highlight the dominant performance factors and demonstrate the feasibility of the approach for a range of payload geometries.

Payload transportation via connected modular unmanned aerial vehicles is an emerging new area that offers unique advantages over other forms of aerial logistics. When considering rigidly attached, modular, vertical lift, unmanned aerial vehicles, differing payloads and vehicle attachment geometries have a significant effect on the composite aircraft's dynamic response during takeoff and stabilization. With no prior knowledge of payload parameters or vehicle attachment geometry, there is no inherent flightworthiness guarantee for a specific connected configuration. Onground flightworthiness determination can be used to ensure acceptable performance during vehicle takeoff or to prescribe changes to the vehicle attachment geometry if necessary. This paper introduces an algorithm to determine flightworthiness while in partial ground contact by estimating the vehicle attachment positions and payload weight. The algorithm uses a probabilistic estimate of vehicle placement about the payload derived through a Bayesian learning technique to generate the necessary data to deterministically estimate the attached vehicles' positions. Following a description of the algorithm, simulation results are presented to illustrate the performance of the algorithm for a variety of modular aircraft configurations.

In-flight estimates of helicopter weight and mass center can be used to improve flight control system performance and inform condition-based maintenance. This paper introduces a two-stage observer design that uses an extended state observer and Kalman filter to produce real-time weight estimates. The proposed algorithm is practical in the sense that it is designed to provide accurate estimates of helicopter weight in the presence of parametric and non-parametric model errors. By decomposing lumped disturbance estimates into a set of perturbation terms and injecting non-parametric corrections into the model, the sensitivity to model error is reduced. Following a description of the proposed algorithm, performance is demonstrated in simulation using the AH-1G helicopter. Under dynamic excitation, the algorithm is shown to converge reliably to the helicopter’s weight even in the presence of significant parametric model discrepancies. This added robustness is reliant on an accurate mass center estimate. Larger model error sensitivity is exhibited in the presence of non-parametric model discrepancies. Several example cases characterize filter performance as a function of the type and degree of model error.