Browse Topic: Education and training
The inconsistency in bearing data distributions under diverse conditions often affects the representations of the faulty data and leads to indistinct decision boundaries and even negative transfer resulted from overlapping class distributions, greatly limiting the accuracy of the diagnosis model. To cope with the challenge, a pseudo-label-guided dual-supervised alignment (PDSA) method is developed for bearing fault diagnosis across diverse operating scenarios in this paper. To address the fixed alignment strategy issue, an adaptive distribution alignment layer is incorporated to ResNet18 to achieve dynamic data distribution alignment under varying condition, To enhance classification performances, a dual-supervised mechanism, comprising shallow-layer supervised contrastive learning is introduced through target domain pseudo-labels in target domain and deep-layer regularization class consistency. Experiments on two publicly available bearing datasets demonstrated this model realizes refined class-level alignment, strengthens fault states representation, and shows notable superiority in both accuracy and robustness.
Regarding the external sling load system of heavy-lift helicopters, the influence of the law of lifting point position on flight control stability characteristics has not been distinctly explained. To address this challenge, this paper constructs a sling load flight simulation model based on multi-body dynamics. Overall, the proposed model consists of four parts, including the rotor aeroelastic coupling model, the fuselage rigid body dynamics model, the flexible sling model, and the slung object rigid body model. Furthermore, through the hub six-degree-of-freedom rigid model and the flexible sling model, this paper realizes the dynamic coupling between the components. On this basis, taking the CH-53E heavy-lift helicopter as the research object, this paper utilizes real flight test data to validate the multi-body dynamic model. Subsequently, this paper systematically analyzes the influence of different lifting points’ lateral position, sling load mode, load-mass ratio, and forward flying speed on helicopter control stability characteristics. Simulation results indicate that the lifting point location exerts a significant impact on the helicopter’s trim attitude angles and dynamic stability. Of them, the lifting point location of the front center of gravity is the optimal in terms of trim characteristics and eigenvalue distribution. Furthermore, within a certain flight speed range, the lifting point of the front center of gravity demonstrates superior speed adaptability and system robustness. Apart from providing a solid theoretical basis for the lifting point layout design of the external sling load system of heavy-lift helicopters, the research results have important engineering application value for improving the safety of sling load flight of heavy-lift helicopters.
When quadrotor unmanned aerial vehicles (UAVs) operate in urban low-altitude airspace, especially within complex environments, their sensor perception signals are highly susceptible to blockages, deviations, and the inclusion of high-frequency noise. These factors, in turn, induce nonlinear variations in the UAVs’ flight mechanical properties, giving rise to abnormal flight stability issues such as attitude jitter, altitude fluctuations, and trajectory deviations. To address these challenges, this paper puts forward a method aimed at enhancing the positional accuracy of quadrotor UAVs, which is based on Extended Kalman Filter (EKF) multi-sensor fusion. In conjunction with the redundant configuration of sensors, a proportional-integral controller is specifically designed to allow optical flow sensors to compensate for the speed data generated by inertial sensors. Building on the EKF method, a comprehensive data fusion model is established, encompassing both position and speed states. Leveraging the MATLAB platform, trajectory flight simulations are conducted, utilizing multi-sensor data fused via EKF, with the sensor suite including GPS, IMU, Optical Flow sensors, and Barometers. The simulation results demonstrate that this proposed method can effectively mitigate the adverse impacts of environmental interference and sensor noise on the positional accuracy of quadrotors. By continuously correcting position information and accurately estimating position states, it significantly improves the UAVs’ flight position accuracy. This research outcome lays a robust and theoretically sound foundation for in-depth investigations on critical issues related to general aviation applications, such as the safe and efficient autonomous flight, adaptive and reliable intelligent navigation, and ultra-precise and mission-critical operations of quadrotor UAVs, thereby significantly contributing to the sustained and innovative advancement of the field.
Rigorous validation of SAE Levels 3 and 4 autonomous systems increasingly relies on simulation. However, the simulation-reality gap remains a challenge for human-in-the-loop assessments. This study empirically quantifies the behavioral fidelity of the Car-Learning-to-Act (CARLA) simulator by recreating specific real-world traffic scenarios using the high-precision exiD drone dataset. Twenty-five participants performed a series of maneuvers, including lane changes and time-critical cut-ins. Their performance was analyzed using Dynamic Time Warping (DTW), driver profiling, and Time-to-Collision (TTC) metrics. The findings reveal a clear distinction between relative and absolute behavioral validity. In strategic decision-making tasks, the simulation demonstrated remarkably high temporal fidelity. DTW analysis explained 94% of the trajectory variance. Participants initiated lane changes with an average lag of -9 frames (0.36 s) compared to naturalistic references. These results indicate that, despite the absence of peripheral optical flow, the simulator successfully elicits temporally correlated decision-making patterns suitable for assessing strategic driver intent. However, physical execution in reactive scenarios revealed significant absolute discrepancies. Although the high Pearson correlation (r ≈ 0.89) in velocity profiles proves that drivers recognize and react to hazards with realistic timing, their physical inputs were exaggerated. Participants displayed digital, over-modulated braking responses and maintained a negative safety bias of -11.26 m, a deviation attributed to the lack of vestibular g-force feedback and geometric minification. Furthermore, distinct driver profiles emerged. Risk-oriented participants exhibited a gaming effect by neglecting safety margins. In conclusion, while CARLA is highly valid for testing the temporal logic of driver interactions, absolute dynamics require calibration functions, such as force-feedback (pedal) tuning and visual deceleration cues like camera shake, to compensate for sensory limitations before it can be used for safety-critical validation.
In recent years, the automotive industry has faced increasing pressure to accelerate development cycles and reduce costs. Simultaneously, ride comfort standards have risen due to the ongoing integration of autonomous driving functionalities. Consequently, it has become essential to ensure that ride comfort attains a high degree of maturity at the very early stages of the automotive development process. This necessitates the establishment of objective criteria that enable the reliable estimation of subjective ride comfort, utilizing simulation-based assessment methods. This study introduces a methodological framework designed to systematically translate the manufacturer specific subjective perception and assessment of ride comfort into objective descriptions using a dynamic driving simulator. The framework is conceived as a generic approach, enabling the comprehensive application to a wide spectrum of subjective ride comfort phenomena, while being specifically optimized for the challenges of the automotive industry. Employing this framework facilitates the derivation of highly detailed, objective descriptions of subjective ride comfort evaluations, which promotes the achievement of advanced ride comfort maturity for new vehicles in early development phases and supports the overall enhancement of ride comfort. The exemplary application of the framework to a transient, one-dimensional ride comfort phenomenon demonstrates its capability to derive robust objective models from subjective evaluations conducted with professional test drivers in a dynamic driving simulator environment.
Realistic seat vibration reproduction is essential for delivering authentic haptic cues and enhancing driver immersion in driving simulators. Unlike direct playback of road recordings, simulator applications require vibration synthesis that responds interactively to driver inputs and vehicle dynamics. Reproducing these vibrations at the seat is often complicated by actuator bandwidth limitations and the dynamic behaviour of the seat structure itself, which can alter the intended target response. This work presents vibration synthesis and seat dynamics compensation strategies implemented on a single-axis seat vibration reproduction system equipped with a vertical actuator. Frequency Response Functions (FRFs) were measured to characterise the system dynamics under single-axis excitation. Run-up and coast-down tests were conducted on the seat and compared to target responses measured on an actual vehicle under operational conditions. Several seat dynamics compensation strategies were evaluated, including inverse FRF filtering, energy-based matching, and manual frequency-domain equalisation. The results indicate that single-input single-output (SISO) compensation approaches can achieve reasonable performance under independent single-axis excitation when compared to alternative methods. The study presents a comparative evaluation of these methodologies and highlights their respective challenges and limitations.
This digital standard is a requirements extract of AS13001A Delegated Product Release Verification Training Requirements. This file contains a general requirements extraction as well as files that are optimized for use with Doors Classic, Siemens Polarian, and PTC.
This study presents a data-driven approach for strengthening aviation safety by integrating human factors assessment with modern predictive modeling techniques. The work focuses on understanding how human performance, operational conditions, and system-level interactions collectively influence safety risk, and how these interactions can be quantified to support improved design and decision-making. Unlike previous studies that address human factors or predictive modeling in isolation, this research offers a unified framework that links causal human factors indicators with statistical modeling, feature extraction, and machine learning based risk estimation. The novelty of this work lies in the structured pipeline that transforms raw categorical and narrative human factors information into measurable predictors that can be analyzed using structural modeling and machine learning. The methodology includes data preparation, dimensionality reduction, latent pattern discovery, dependence modeling, model training, and interpretability analysis. The study demonstrates how this pipeline uncovers hidden relationships among operational errors, environmental influences, maintenance actions, design considerations, and crew behavior. The findings show that the integrated approach improves the accuracy and stability of risk prediction and highlights specific human factors patterns that consistently contribute to elevated risk levels. These insights support targeted mitigation strategies, inform design improvements, and help prioritize safety interventions. The work concludes that a combined human factors and predictive modeling framework enhances the ability of organizations to identify vulnerabilities earlier, allocate resources more effectively, and strengthen system resilience. This approach is adaptable to diverse aviation contexts and offers a practical path for transforming human factors data into actionable safety intelligence.
The U.S. ARMY Primary Helicopter Center/School, USAPHC/S, was activated at Fort Wolters on September 26, 1956. Located in north-central Texas, the school would train over 40,000 helicopter pilots during 17 years of operation, through the end of the Vietnam War in 1973. Approximately 95 percent of all helicopter pilots who flew in Vietnam would pass through Wolters. Students included active-duty Army Officers, Warrant Officer Candidates, and Officers representing 33 allied countries. They trained for 16 weeks at Wolters and then another 16 weeks of advanced training at Fort Rucker, Alabama before earning army aviator wings. At the peak of activity in 1968, Wolters was sending 608 pilots per month to Fort Rucker. Students flew a total of 1,285 piston-powered OH-13, OH-23D, and TH-55A training helicopters departing out of three different heliports. It is a mystical place that still lives in the history of Army Aviation through the helicopter pilots who trained there. This is their story.
The FAA VR-HeliSTART (Virtual Reality-Helicopter Simulator Training for Airplane to Rotorcraft Transition) is a 15-week study conducted at Marshall University (WV) to determine the effectiveness of an H125 VR reduced-motion platform simulator in training fixed-wing pilots to fly helicopters. Eleven students received three four-week blocks of instruction in the flight simulator, each followed by a simulator evaluation and a helicopter evaluation. This paper presents results for eleven hovering maneuvers trained and evaluated in the study. The evaluation of the students relied on both an objective and a subjective evaluation: a flight parameter analysis against Airman Certification Standards criteria, and an assessment by certified flight instructors. A key finding is that simulator training enabled all pilots to perform most hover maneuvers on their first helicopter flight without intervention, although sometimes below standards. Overall, results also suggest that while the simulator provides a useful learning environment for basic hover control, the further refinement of the core hovering skills acquired in the simulator did not appear to transfer effectively to the actual helicopter within the time frame of the study. Therefore, this indicates that initial hover training in the simulator is beneficial, but additional improvements still seem to require practice in the actual helicopter.
This paper presents a spatio-temporal graph neural network (STGNN) centric approach to enable heterogeneous agents to collaborate and cooperate for different types of missions. The STGNN-centric approach and corresponding autonomy are encapsulated in the Advanced Graph-enabled Network Technology for Collaborative Autonomous Agents (AGENTCA) technology. Various decentralized and distributed control architectures are reported in the literature, but in some instances these approaches do not leverage the inherent graph network which can increase scalability to larger teams and algorithmic efficiency. Specifically, in this paper advances in artificial intelligence are leveraged to parameterize and encode optimal, or nearly optimal, swarm control techniques. For this work, the team focused on developing a diffusion-based STGNN swarm controller using imitation learning. An expert, centralized swarm control law was used to guide the STGNN during the learning process. The STGNN controller enables the swarm to follow a leader while avoiding static and dynamic obstacles and maintaining a desired separation distance from neighbors and obstacles. The approach is demonstrated in simulation with hundreds of agents and in flight tests with up to thirteen test vehicles.
Historical rotor designs for Earth and Mars have typically landed at thrust-weighted solidities of ∼0.1-0.15 as a best compromise of performance and weight. Comprehensive analysis predicts that high solidity rotor designs of more than twice this range have the potential to significantly increase the lift capability of future Mars explorers severely limited by packaging and weight. However, there is limited existing experimental data of high solidity rotor designs at representative densities to quantify the efficiency impact and verify models of the aerodynamic environment. Therefore, the Mars Exploration Program (MEP) funded a joint test campaign between NASA's Jet Propulsion Laboratory, NASA Ames Research Center, and AeroVironment, Inc.to validate performance predictions for low- and high- solidity rotor variants at Mars pressures. Experimental setup, test matrix, data processing, data quality, and performance results for the High Solidity Test (HST) campaign are presented and discussed.
This paper investigates the use of full-body vibrotactile cueing to augment operator perception during swarm teleoperation tasks. Piloted simulations are conducted in a virtual reality (VR) flight simulation environment using a quadcopter swarm model and a nonlinear dynamic inversion (NDI) flight control architecture. A scaled version of the ADS-33 slalom Mission Task Element (MTE) is implemented to evaluate swarm formation maintenance and obstacle avoidance under four experimental conditions: Good Visual Environment (GVE), Degraded Visual Environment (DVE), and each of these conditions augmented with haptic feedback. Haptic cues are delivered through vibrotactile vests and sleeves to convey information on formation deformation and gate proximity. Experimental results involving human participants indicate that haptic feedback improves formation maintenance and increases operators’ situational awareness of follower drone positions without increasing perceived mental workload. While haptic cues provided modest assistance in gate localization, visual conditions remained the dominant factor influencing obstacle avoidance performance. Overall, the results indicate that full-body haptic feedback provides an effective modality for augmenting operator perception and supporting swarm supervision tasks, particularly in visually degraded environments.
The rapid expansion of electric aviation and eVTOL operations introduces tightly coupled challenges related to energy‑constrained aircraft design, battery and thermal management, mission planning, and the generation of certification‑relevant evidence. This paper presents an integrated simulation workflow developed by AVL, Unisphere, and blueflite that combines high‑fidelity electric powertrain and battery models with a guidance‑level, digital‑twin‑based 4‑D trajectory simulation driven by historical weather and operational constraints. At each mission time step, the trajectory layer provides time‑resolved environmental and routing conditions, while the system‑level models compute instantaneous power demand, state‑of‑charge evolution, and thermal response, enabling mission feasibility assessment under realistic wind, temperature, and airspace effects. The workflow is calibrated and validated using flight telemetry from blueflite's active eVTOL cargo aircraft development, ensuring alignment between simulation assumptions and real‑world mission execution. The validated framework is subsequently applied to seasonal route studies and large‑scale virtual flight campaigns spanning multiple regions and years, enabling statistically robust assessment of energy margins, thermal behavior, and mission‑duration variability. The results demonstrate how integrated, traceable simulation can bridge conceptual design and real‑world electric flight operations, supporting informed decision‑making by OEMs and operators in aircraft design, validation, and deployment planning.
Flight simulations are critical for aerial firefighting training, but realistic modelling of aircraft-atmosphere interactions within fire scenarios is particularly challenging. To this end, a two-way-coupled flight simulation system, the Daedalus I framework, has been developed at the University of Glasgow for helicopter firefighting research applications. This paper presents the initial results from flight experiments conducted with different coupling schemes between the rotorcraft model and the GPU-accelerated Lattice Boltzmann atmosphere model within the system. The two-way coupling scheme was first validated using an isolated, transient rotor case. To quantify differences in pilot control and strategy between the two-way, fully-coupled rotor-atmosphere method and two (2) one-way, superposition-based coupling methods, a series of flight experiments were conducted using the bimodal modification of the McRuer pilot model representing human pilot controls, in conjunction with objective performance metrics. The results highlighted noticeable changes in helicopter behaviours and pilot responses across the three tested coupling methods, with the two-way coupling method showing the most resistance to the generated fire disturbance.
Pilot compensation — the effort required to maintain task performance in the face of deficient vehicle characteristics, as rated on the Cooper–Harper Handling Quality Rating (HQR) scale – is the task-performance-anchored measure of workload. While it has traditionally been inferred from control activity alone, recent work shows that eye-movement activity carries complementary information: as compensation rises, control inputs increase while visual scanning narrows, so neither channel alone captures the full picture. This paper proposes the pilot action metric, which combines control-stick and eye-movement activity rates so that both channel responses reinforce the compensation signal. A shared-slope regression model with per-pilot intercepts is evaluated via leave-one-out cross-validation on 16 simulator runs flown by three military test pilots across four mission task elements. The combined metric succeeds where either channel alone fails, reproducing 94% of ratings to within ±1 HQR. The model further yields a conservative maximum-tolerable-compensation boundary that is consistent with independently derived flight-test data.
This paper introduces a robust supervised machine learning framework for estimating helicopter gross weight during the takeoff phase. The methodology leverages high-fidelity datasets from Airbus's global in-service fleet to ensure a reliable training foundation. At the core of the approach is a long short-term memory recurrent neural network, supported by a patented data-curation pipeline designed to maintain high data integrity. To align with rigorous aviation safety standards, the study outlines a learning assurance process compliant with EASA guidelines, specifically addressing safety assessment objectives for machine learning. A central innovation is the characterization and monitoring of the model's operational design domain through multidimensional functional principal component analysis. By projecting high-dimensional, non-linear sensor data into a manageable tabular subspace, this approach enables the definition of safety envelopes using explainable and efficient classical methods. Validated against diverse real-world flight profiles, the framework demonstrates high predictive accuracy, marking a significant milestone toward deploying the model on airborne targets for safety-critical functions such as condition-based maintenance.
The FAA VR-HeliSTART (Virtual Reality-Helicopter Simulator Training for Airplane to Rotorcraft Transition) is a 15-week study conducted at Marshall University (WV) to determine the effectiveness of an H125 VR reduced-motion platform simulator in training fixed-wing pilots to fly helicopters. 11 students received three four-week blocks of instruction from certified flight instructors in the flight simulator, each followed by evaluations in both the simulator and an actual H125 helicopter, covering 36 maneuvers drawn from the commercial helicopter Airman Certification Standards. A mixed-methods approach combined objective flight parameter analysis with subjective assessments from evaluators, instructors, and students. Results indicate broadly positive transfer of training, with students demonstrating at least private pilot level performance on 70% or more of maneuvers on their first helicopter flight, and consistent improvement across subsequent evaluations. However, specific areas of negative or limited transfer were identified, most notably Vortex Ring State recovery, approaches, and hover work, driven by limitations of the head-mounted display and motion platform. This paper presents the methodology and overall results, with future papers addressing individual maneuver groups in greater detail.
Prior work demonstrated that acceleration washout in motion simulators produces decay-rate sensing ambiguity within the vestibular system, forcing pilots to rely on visual cues for control. While Pilot Induced Oscillation Ratings (PIORs) for flight and simulation have been matched using different sensing thresholds, a quantitative basis for the 50% reduction in the visual decay-rate threshold has remained elusive. This paper provides evidence that pilots perceive decay rate proprioceptively through stick force during both flight and simulation, rather than through vestibular or visual channels. The residues of the stick-force sensitivity transfer function reflect the amplification or attenuation of neighboring zeros and poles; when these residues fall outside the human's 30 dB tactile sensory window, the resulting decay rate becomes imperceptible. Modeling reveals that stabilization via the visual channel in simulators produces dominant mode characteristics - decay rates, frequencies, and residues - that diverge significantly from vestibular-stabilized flight. The Virtual Vestibular Cueing (VVC) technique is introduced to tune the visual channel using pseudo-vestibular rate cueing, optimally aligning the simulated dominant mode with flight-validated residues and decay rates. A preliminary fixed-base study demonstrates that VVC synchronizes performance and subjective ratings by restoring this tactile-vestibular harmony. This work establishes that handling qualities are fundamentally an Information Theory problem: Level 1 ratings occur when the residue-to-decay gradient is tuned to map the dominant mode's physical 'elbow' onto the human sensory window. This paper establishes that perceptual fidelity is not determined by the closed-loop vehicle residues, but by the force sensitivity residues. By identifying the stick-force channel as the superior conduit for this mapping, VVC allows designers to restore informational integrity in virtual environments across all axes of aircraft dynamics. VVC ensures that the visual channel supports a control strategy where the tactile feedback remains within the human sensory window, preventing the 'wrong reading' that leads to simulator-to-flight rating mismatches.
This study investigates the post-failure flight dynamics of a 1200 lb classical octocopter under single motor inoperative condition using nonlinear time-domain simulations with a baseline feedback controller. A physics based propulsion sizing strategy is developed using IEC duty cycle definitions where continuous requirements are derived from nominal hover with margin and short time capability is used to accommodate elevated post failure loads. The selected motor satisfies both regimes and enables transient overdrive without excessive weight penalty. Simulation results in hover and forward flight at the best range speed showing that the vehicle can recover from any single motor failure and retrim using inherent redundancy without fault identification. However, recovery involves significant transient attitude excursions and altitude loss, and requires substantial increases in motor power, with multiple motors exceeding S1 power limits. Post-failure maneuver simulations indicate retained controllability with some degradation and increased coupling. These simulations demonstrate that the proposed motor sizing enables necessary operation post-failure while avoiding unnecessary oversizing.
This study evaluates the operational impact of multiple concurrent spatialized auditory cues during high-workload rotorcraft missions. A controlled, within-subject flight simulation experiment was conducted in which military-qualified rotorcraft pilots completed continuous multi-objective missions including formation flying, visual asset detection, collision avoidance, and emergency landing tasks. Each mission was flown under spatialized (3D) and non-spatialized (2D) audio rendering conditions while cue composition remained constant. Preliminary results indicate that under complex, formation-dominant workload conditions, pilots consistently prioritized visually anchored tasks and largely deprioritized auditory cue information regardless of spatial rendering. Collision avoidance cues did not produce observable evasive responses, and reported cue trust remained low without prior training. Although limited performance improvements were observed in isolated conditions, participants reported consciously suppressing audio cues. These findings suggest that effective integration of spatial audio requires structured training, procedural embedding, and deliberate workload redistribution rather than perceptual enhancement alone.
Welcome, once again, to our annual digital-only issue. We remain committed, of course, to our eight print issues each year, but letting the benefits of digital media come to the fore is always its own fun challenge. No matter what your role is, you should find something engaging and educational among these virtual pages. We've got videos, animations and other multimedia components for you to click on and learn from in this issue. My personal favorite example in here of a video replacing a thousand words can be found in our cover story. It's not like the text is indecipherable, but simply seeing how adaptive headlight beams adjust where they throw their high beams so as to not blind other drivers in almost all situations is just cool. Headlight technology has come a long, long way in the last decade, from the low lights of old helping clear a path home to today's often-annoying high beams that may provide extra safety to the driver, but make the blood of oncoming drivers run dry.
Autonomous vehicle navigation requires accurate prediction of driving path curvature to ensure smooth and safe trajectory planning. This paper presents a novel approach to curvature prediction using deep neural networks trained on GPS-derived ground truth data, rather than model predictions, providing a more accurate training signal that reflects actual vehicle motion. We develop a multi-modal neural network architecture with temporal GRU encoders that processes vision features, driver intent signals, historical curvature, and vehicle state parameters to predict curvature. A key innovation is the use of GPS-based actual curvature measurements computed from vehicle motion data (κ = ωz/v) as training supervision, enabling the model to learn from real-world driving patterns. The model is trained on 5,322 samples from real-world driving data collected on The University of Oklahoma’s Norman Campus using a Comma 3X device and a 2025 Nissan Leaf electric vehicle. Experimental results demonstrate high steering curvature prediction accuracy with a Pearson correlation coefficient of 0.805, Mean Absolute Error of 0.027654, and Root Mean Squared Error of 0.034402 on the validation set. The model achieves stable convergence within 10 epochs and maintains consistent performance across diverse driving scenarios, from straight highway segments to complex turning maneuvers. This work contributes to autonomous driving technology by demonstrating the effectiveness of GPS-supervised learning for curvature prediction, successfully deployed in OpenPilot’s production system with real-time inference at 5 Hz.
Items per page:
50
1 – 50 of 6560