Search for: All records

Uncrewed aerial vehicles are integral to a smart city framework, but the dynamic environments above and within urban settings are dangerous for autonomous flight. Wind gusts caused by the uneven landscape jeopardize safe and effective aircraft operation. Birds rapidly reject gusts by changing their wing shape, but current gust alleviation methods for aircraft still use discrete control surfaces. Additionally, modern gust alleviation controllers challenge small uncrewed aerial vehicle power constraints by relying on extensive sensing networks and computationally expensive modeling. Here we show end-to-end deep reinforcement learning forgoing state inference to efficiently alleviate gusts on a smart material camber-morphing wing. In a series of wind tunnel gust experiments at the University of Michigan, trained controllers reduced gust impact by 84% from on-board pressure signals. Notably, gust alleviation using signals from only three pressure taps was statistically indistinguishable from using six pressure tap signals. By efficiently rejecting environmental perturbations, reduced-sensor fly-by-feel controllers open the door to small uncrewed aerial vehicle missions in cities.

 

Despite its prevalence in neurosensory systems for pattern recognition, event detection, and learning, the effects of sensory adaptation (SA) are not explored in reservoir computing (RC). Monazomycin‐based biomolecular synapse (MzBS) devices that exhibit volatile memristance and short‐term plasticity with two strength‐dependent modes of response are studied: facilitation and facilitation‐then‐depression (i.e., SA). Their ability to perform RC tasks including digit recognition, nonlinear function learning, and aerodynamic gust classification via combination of model‐based device simulations and physical experiments where SA presence is controlled is studied. Simulations exhibiting moderate SA achieve significantly higher accuracy classifying a custom 5 × 5 binary digit set, with experimental validation achieving maximum testing accuracies of 90%. Classifications of the Modified National Institute of Standards and Technology (MNIST) handwritten digit dataset achieve a maximum testing accuracy of 94.34% in devices with SA. Fitting error of the Mackey–Glass time series is also significantly reduced by SA. Experimentally obtained pressure distributions representing gusts on an airfoil in a wind tunnel are classified by MzBS reservoirs. Reservoirs exhibiting SA achieve 100% accuracy, unlike MzBS reservoirs without SA and comparable static neural networks.

 

Some bird species fly inverted, or whiffle, to lose altitude. Inverted flight twists the primary flight feathers, creating gaps along the wing’s trailing edge and decreasing lift. It is speculated that feather rotation-inspired gaps could be used as control surfaces on uncrewed aerial vehicles (UAVs). When implemented on one semi-span of a UAV wing, the gaps produce roll due to the asymmetric lift distribution. However, the understanding of the fluid mechanics and actuation requirements of this novel gapped wing were rudimentary. Here, we use a commercial computational fluid dynamics solver to model a gapped wing, compare its analytically estimated work requirements to an aileron, and identify the impacts of key aerodynamic mechanisms. An experimental validation shows that the results agree well with previous findings. We also find that the gaps re-energize the boundary layer over the suction side of the trailing edge, delaying stall of the gapped wing. Further, the gaps produce vortices distributed along the wingspan. This vortex behavior creates a beneficial lift distribution that produces comparable roll and less yaw than the aileron. The gap vortices also inform the change in the control surface’s roll effectiveness across angle of attack. Finally, the flow within a gap recirculates and creates negative pressure coefficients on the majority of the gap face. The result is a suction force on the gap face that increases with angle of attack and requires work to hold the gaps open. Overall, the gapped wing requires higher actuation work than the aileron at low rolling moment coefficients. However, above rolling moment coefficients of 0.0182, the gapped wing requires less work and ultimately produces a higher maximum rolling moment coefficient. Despite the variable control effectiveness, the data suggest that the gapped wing could be a useful roll control surface for energy-constrained UAVs at high lift coefficients.

 

Birds perform astounding aerial maneuvers by actuating their shoulder, elbow, and wrist joints to morph their wing shape. This maneuverability is desirable for similar-sized uncrewed aerial vehicles (UAVs) and can be analyzed through the lens of dynamic flight stability. Quantifying avian dynamic stability is challenging as it is dictated by aerodynamics and inertia, which must both account for birds’ complex and variable morphology. To date, avian dynamic stability across flight conditions remains largely unknown. Here, we fill this gap by quantifying how a gull can use wing morphing to adjust its longitudinal dynamic response. We found that it was necessary to adjust the shoulder angle to achieve trimmed flight and that most trimmed configurations were longitudinally stable except for configurations with high wrist angles. Our results showed that as flight speed increases, the gull could fold or sweep its wings backward to trim. Further, a trimmed gull can use its wing joints to control the frequencies and damping ratios of the longitudinal oscillatory modes. We found a more damped phugoid mode than similar-sized UAVs, possibly reducing speed sensitivity to perturbations, such as gusts. Although most configurations had controllable short-period flying qualities, the heavily damped phugoid mode indicates a sluggish response to control inputs, which may be overcome while maneuvering by morphing into an unstable flight configuration. Our study shows that gulls use their shoulder, wrist, and elbow joints to negotiate trade-offs in stability and control and points the way forward for designing UAVs with avian-like maneuverability. 

Abstract Some bird species exhibit a flight behavior known as whiffling, in which the bird flies upside-down during landing, predator evasion, or courtship displays. Flying inverted causes the flight feathers to twist, creating gaps in the wing’s trailing edge. It has been suggested that these gaps decrease lift at a potentially lower energy cost, enabling the bird to maneuver and rapidly descend. Thus, avian whiffling has parallels to an uncrewed aerial vehicle (UAV) using spoilers for rapid descent and ailerons for roll control. However, while whiffling has been previously described in the biological literature, it has yet to directly inspire aerodynamic design. In the current research, we investigated if gaps in a wing’s trailing edge, similar to those caused by feather rotation during whiffling, could provide an effective mechanism for UAV control, particularly rapid descent and banking. To address this question, we performed a wind tunnel test of 3D printed wings with a varying amount of trailing edge gaps and compared the lift and rolling moment coefficients generated by the gapped wings to a traditional spoiler and aileron. Next, we used an analytical analysis to estimate the force and work required to actuate gaps, spoiler, and aileron. Our results showed that gapped wings did not reduce lift as much as a spoiler and required more work. However, we found that at high angles of attack, the gapped wings produced rolling moment coefficients equivalent to upwards aileron deflections of up to 32.7° while requiring substantially less actuation force and work. Thus, while the gapped wings did not provide a noticeable benefit over spoilers for rapid descent, a whiffling-inspired control surface could provide an effective alternative to ailerons for roll control. These findings suggest a novel control mechanism that may be advantageous for small fixed-wing UAVs, particularly energy-constrained aircraft. 

Abstract Forpractical considerations reinforcement learning has proven to be a difficult task outside of simulation when applied to a physical experiment. Here we derive an optional approach to model free reinforcement learning, achieved entirely online, through careful experimental design and algorithmic decision making. We design a reinforcement learning scheme to implement traditionally episodic algorithms for an unstable 1-dimensional mechanical environment. The training scheme is completely autonomous, requiring no human to be present throughout the learning process. We show that the pseudo-episodic technique allows for additional learning updates with off-policy actor-critic and experience replay methods. We show that including these additional updates between periods of traditional training episodes can improve speed and consistency of learning. Furthermore, we validate the procedure in experimental hardware. In the physical environment, several algorithm variants learned rapidly, each surpassing baseline maximum reward. The algorithms in this research are model free and use only information obtained by an onboard sensor during training. 

Smooth camber morphing aircraft offer increased control authority and improved aerodynamic efficiency. Smart material actuators have become a popular driving force for shape changes, capable of adhering to weight and size constraints and allowing for simplicity in mechanical design. As a step towards creating uncrewed aerial vehicles (UAVs) capable of autonomously responding to flow conditions, this work examines a multifunctional morphing airfoil’s ability to follow commands in various flows. We integrated an airfoil with a morphing trailing edge consisting of an antagonistic pair of macro fiber composites (MFCs), serving as both skin and actuator, and internal piezoelectric flex sensors to form a closed loop composite system. Closed loop feedback control is necessary to accurately follow deflection commands due to the hysteretic behavior of MFCs. Here we used a deep reinforcement learning algorithm, Proximal Policy Optimization, to control the morphing airfoil. Two neural controllers were trained in a simulation developed through time series modeling on long short-term memory recurrent neural networks. The learned controllers were then tested on the composite wing using two state inference methods in still air and in a wind tunnel at various flow speeds. We compared the performance of our neural controllers to one using traditional position-derivative feedback control methods. Our experimental results validate that the autonomous neural controllers were faster and more accurate than traditional methods. This research shows that deep learning methods can overcome common obstacles for achieving sufficient modeling and control when implementing smart composite actuators in an autonomous aerospace environment.

 

Structural health monitoring of fiber reinforced composites is an extensive field of research that aims to reduce maintenance costs through in-situ damage detection. However, the need for externally bonded sensor systems and complicated fabrication processes limit the widespread application of most current structural health monitoring techniques. This work introduces a novel multifunctional fiber reinforced composite that relies on a ferroelectric prepreg fabricated using dehydrofluorinated (DHF) polyvinylidene fluoride (PVDF), which exhibits a thermally stable piezoelectric response. The self-sensing material presented in this work requires minimal external components, as the piezoelectric sensing mechanism is fully contained within the composite. This is accomplished by fabricating a ferroelectric prepreg consisting of DHF PVDF infused woven fiberglass, which is sandwiched between woven carbon fabric layers that act as electrodes, thus forming a piezoelectric sensor fabricated with entirely structural composite materials. Notably, the sensing material is a fully distributed prepreg rather than discretely embedded sensors which enables simplified monitoring of complex structures. As the composite experiences damage under flexural and tensile loading, the internal change in strain results in a charge separation that is detectable as a voltage emission across the sample electrodes. The self-sensing capabilities of this material are explored using traditional mechanical testing techniques, showing comparable performance to common damage detection methods, all while eliminating the need for external bonding of sensors to the structure.