University of Toronto Institute for Aerospace Studies (UTIAS)

Permanent URI for this communityhttps://hdl.handle.net/1807/69352

UTIAS is a graduate studies and research institute, forming part of the Faculty of Applied Science and Engineering at the University of Toronto .

The continuing involvement of the University of Toronto in the aerospace sciences since World War I eventually led to the establishment of this institute in 1949. Originally called the Institute of Aerophysics, it is now known as the Institute for Aerospace Studies.

Browse

Now showing 1 - 20 of 38

Airfoil Optimization Using Practical Aerodynamic Design Requirements
(American Institute of Aeronautics and Astronautics, 2010-09-01) Buckley, Howard P.; Zhou, Beckett Y.; Zingg, David W.
Practical aerodynamic design problems must balance the goal of performance optimization over a range of on-design operating conditions with the need to meet design constraints at various off-design operating conditions. Such design problems can be cast as multipoint optimization problems where the on-design and off-design operating conditions are represented as design points with corresponding objective/constraint functions. Two methods are presented for obtaining optimal airfoil designs that satisfy all design objectives and constraints. The first method uses an unconstrained optimization algorithm where the optimal design is achieved by minimizing a weighted sum of the objective functions at each of the operating conditions. To address the competing design objectives between on-design and off-design operating conditions, an automated procedure is used to efficiently weight the off-design objective functions so as to limit their influence on the overall optimization while satisfying the design constraints. The second method uses the constrained optimization algorithm SNOPT, which allows the aerodynamic constraints imposed at the off-design operating conditions to be treated explicitly. Both methods are applied to the design of an airfoil for a hypothetical aircraft where the problem is formulated as an 18-point multipoint optimization.
Auto-Ignition and Flame Stabilization of Hydrogen / Natural Gas / Nitrogen Jets in a Vitiated Cross-Flow at Elevated Pressure
(Elsevier, 2013) Steinberg, Adam M.; Fleck, Julia M.; Griebel, Peter; Arndt, Christoph M.; Aigner, Manfred
The influence of natural gas (NG) on the auto-ignition behavior of hydrogen (H2)/nitrogen (N2) fuel jets injected into a vitiated cross-flow was studied at conditions relevant for practical combustion systems (p = 15 bar, Tcross-flow = 1173 K). In addition, the flame stabilization process following auto-ignition was investigated by means of high-speed luminosity and shadowgraph imaging. The experiments were carried out in an optically accessible jet-in-cross-flow (JICF) test section. In a H2/NG/N2 fuel mixture, the fraction of H2 was stepwise increased while keeping the N2 fraction approximately constant. Two different jet penetration depths, represented by two N2 fraction levels, were investigated. The results reveal that auto-ignition kernels occurred even for the lowest tested H2 fuel fraction (XH2/NG = XH2 / (XH2 + XNG) = 80%), but did not initiate a stable flame in the duct. Increasing XH2/NG decreased the distance between the initial position of the auto-ignition kernels and the fuel injector, finally leading to flame stabilization. The H2 fraction for which flame stabilization was initiated depended on jet penetration; flame stabilization occurred at lower H2 fractions for the higher jet penetration depth (XH2/NG = 91% compared to 96%), revealing the influence of different flow fields and mixing characteristics on the flame stabilization process. It is hypothesized that the flame stabilization process is related to kernels extending over the duct height and thus altering the upstream conditions due to considerable heat release. This enabled subsequent kernels to occur close to the fuel injector until they could finally stabilize in the recirculation zone of the jet lee.
Autoignition of Hydrogen / Nitrogen Jets in Vitiated Air Crossflows at Different Pressures
(Elsevier, 2013) Steinberg, Adam M.; Fleck, Julia M.; Griebel, Peter; Arndt, Christoph M.; Naumann, Clemens; Aigner, Manfred
Autoignition of hydrogen/nitrogen jets in crossflows of vitiated air has been experimentally studied at conditions relevant for gas turbine combustor operation, including practical pressures (5, 10, 15 bar), crossflow temperatures (Tcf = 1185 K and 1143 K), crossflow velocities (ucf = 200 and 300 m/s), and jet momentum ranges. Experiments were performed in an optically accessible duct, in which the appearance of autoignition events was investigated using high-speed imaging. At the different crossflow parameters, the H2 mole fraction XH2 in the fuel jet was incrementally increased until autoignition created a stable flame. During this process, the autoignition kernel and subsequent flame dynamics were recorded at a rate of 30 kHz. In order to compare the experimental autoignition characteristics of the partially premixed, turbulent system to a homogeneous system, kinetic simulations of homogenous systems were conducted for pressures, temperatures, and vitiated air compositions corresponding to the experiments. The ignition delay times estimated from the experiment were considerably shorter than those from the kinetic simulations, and exhibited a different pressure dependency. These differences suggest that, in the current flow configuration, autoignition is strongly affected by turbulent mixing and flow field characteristics.
Batch nonlinear continuous-time trajectory estimation as exactly sparse Gaussian process regression
(Springer Verlag, 2015) Anderson, Sean; Barfoot, Timothy D.; Tong, Chi Hay; Särkkä, Simo
In this paper, we revisit batch state estimation through the lens of Gaussian process (GP) regression. We consider continuous-discrete estimation problems wherein a trajectory is viewed as a one-dimensional GP, with time as the independent variable. Our continuous-time prior can be defined by any nonlinear, time-varying stochastic differential equation driven by white noise; this allows the possibility of smoothing our trajectory estimates using a variety of vehicle dynamics models (e.g., ‘constant-velocity’). We show that this class of prior results in an inverse kernel matrix (i.e., covariance matrix between all pairs of measurement times) that is exactly sparse (block-tridiagonal) and that this can be exploited to carry out GP regression (and interpolation) very efficiently. When the prior is based on a linear, time-varying stochastic differential equation and the measurement model is also linear, this GP approach is equivalent to classical, discrete-time smoothing (at the measurement times); when a nonlinearity is present, we iterate over the whole trajectory to maximize accuracy. We test the approach experimentally on a simultaneous trajectory estimation and mapping problem using a mobile robot dataset.
Collapse mechanisms of sandwich beams with composite faces and a foam core, loaded in three-point bending. Part I: analytical models and minimum weight design
(Elsevier, 2004-04) Steeves, Craig A; Fleck, Norman A
Analytical predictions are made for the three-point bending collapse strength of sandwich beams with composite faces and polymer foam cores. Failure is by the competing modes of face sheet microbuckling, plastic shear of the core, and face sheet indentation beneath the loading rollers. Particular attention is paid to the development of an indentation model for elastic faces and an elastic-plastic core. Failure mechanism maps have been constructed to reveal the operative collapse mode as a function of geometry of sandwich beam, and minimum weight designs have been obtained as a function of an appropriate structural load index. It is shown that the optimal designs for composite-polymer foam sandwich beams are of comparable weight to sandwich beams with metallic faces and a metallic foam core.
Collapse mechanisms of sandwich beams with composite faces and a foam core, loaded in three-point bending. Part II: experimental investigation and numerical modelling
(Elsevier, 2004-04) Steeves, Craig A; Fleck, Norman A
This study focuses on the competing collapse mechanisms for simply supported sandwich beams with composite faces and a PVC foam core subjected to three point bending. The faces comprise Hexcel Fibredux 7781-914G woven glass fibre-epoxy prepreg, while the core comprises closed cell Divinycell PVC foam of relative density 6.6% and 13.3%. The mechanical properties of the face sheets and core are measured independently. Depending upon the geometry of the beam and the relative properties of the constituents, collapse is by core shear, face sheet microbuckling or by indentation beneath the middle loading roller. A systematic series of experiments and finite element simulations have been performed in order to assess the accuracy of simple analytic expressions for the strength. In general, the analytic expressions for peak load are adequate; however, simple beam theory becomes inappropriate and the analytic models are inaccurate for stubby beams with thick faces relative to the core thickness. A failure mechanism map is constructed to reveal the dependence of the dominant collapse mechanism upon the geometry of the beam.
Concepts for structurally robust materials that combine low thermal expansion with high stiffness
(Elsevier, 2007-09) Steeves, Craig A; dos Santos e Lucato, Sergio L; He, Ming; Antinucci, Emilio; Hutchinson, John W; Evans, Anthony G
A family of robust stretch-dominated bimaterial lattices is introduced which combines low (or zero) thermal expansion with high stiffness, structural robustness over wide temperature ranges and manufacturing facility. This combination of properties is unavailable through any other material solution. The concept uses two constituents configured as adjoining sub-lattices. It accommodates the thermal expansion through rotation of the members of one sub-lattice. Moreover, the lattice exhibits large stiffness to weight because it is fully triangulated and does not rely on rotational resistance at the joints for structural rigidity. A wide range of constituents can be used to build the new lattices enabling many desirable properties to be incorporated, especially high strength and toughness. Examples of both planar and volumetric lattices are presented, and their thermo-mechanical properties derived. The results are verified by conducting experiments and finite element simulations on a lattice fabricated using aluminium and titanium alloy constituents.
Coupled dynamics of lift-off and precessing vortex core formation in swirl flames
(Elsevier, 2016) An, Qiang; Kwong, Wing Yin; Geraedts, Benjamin D.; Steinberg, Adam M.
The lift-off mechanism of swirl stabilized premixed flames was investigated using high repetition rate OH planar laser induced fluorescence (PLIF), particle image velocimetry (PIV), and OH* chemiluminescence. At steady operating conditions, the studied flames stochastically transitioned between attached and lifted configurations, with an increasing percentage of time spent in the detached state with increasing velocity. Neither the percentage of time spent in the detached state nor the final detachment conditions were predicted by the Damköhler number. The lift-off process involved several coupled phenomena, namely local flame extinction near the flame base, development of a helical precessing vortex core (PVC), and eventual total extinction of the flame base. Wavelet analysis demonstrated the correspondence between the flame lift-off height and the strength of the PVC. Furthermore, it showed that the initial local extinction preceded the PVC formation. Hence, the lift-off mechanism involved a stochastic local extinction event near the flame base altering the combustor density field and promoting formation of the PVC. The PVC increased the strain-rate on the flame base, which eventually led to total detachment.
Development and evaluation of gappy-POD as a data reconstruction technique for noisy PIV measurements in gas turbine combustors
(Springer Verlag, 2016) Saini, Pankaj; Arndt, Christoph M.; Steinberg, Adam M.
Low signal-to-noise in particle image velocimetry (PIV) measurements in systems such as high pressure gas turbine combustors can result in significant data gaps that negatively affect subsequent analysis. Here, gappy proper orthogonal decomposition (GPOD) is evaluated as a method of filling such missing data. Four GPOD methods are studied, including a new method that utilizes a median filter (MF) to adaptively select whether a local missing data point is updated after each iteration. These methods also are compared against local Kriging interpolation. The GPOD methods are tested using PIV data without missing vectors that were obtained in atmospheric pressure swirl flames. Parameters studied include the turbulence intensity, amount of missing data, and the amount of noise in the valid data. Two criteria to check for GPOD convergence also were investigated. The MF method filled in the missing data with the lowest error across all parameters tested, with approximately one-third the computational cost of Kriging. Furthermore, the accuracy of MF GPOD was relatively insensitive to the quality of the convergence criterion. Therefore, compared to the three other GPOD methods and Kriging interpolation, the MF GPOD method is an effective method for filling missing data in PIV measurements in the studied gas turbine combustor flows.
Effect of B3+ - N3- on YAG:Dy Thermographic on YAG:Dy Thermographic Phosphor Luminescence
(Optical Society of America, 2014) Kwong, Wing Yin; Steinberg, Adam M.; Chin, Ya Huei
The use of thermographic phosphors for high-temperature (> 1000 K) thermometry currently is limited by loss of signal due to thermal quenching. This work demonstrates a new phosphor generated by substituting tetrahedral site Al3+-O2- in YAG:Dy with B3+-N3- to produce YABNG:Dy. Conventional YAG:Dy and YABNG:Dy phosphors were synthesized using identical sol-gel synthesis techniques. X-ray diffraction measurements showed that both had nearly pure crystalline phases, with a minor secondary yttrium-aluminum-monoclinic (YAM) phase present in the YABNG:Dy. The YABNG:Dy sample had a larger and more spherical primary grain than the YAG:Dy in scanning electron microscopy images. Tests of the thermal response showed that the YABNG:Dy had much stronger phosphorescence emissions than YAG:Dy, likely due to the morphological differences. Furthermore, the onset of thermal quenching was delayed by approximately 100 K for YABGN:Dy compared to YAG:Dy, and the rate of signal decrease with temperature was reduced. This resulted in greater signal-to-noise ratios and less uncertainty in the temperature measurements, particularly at high temperatures.
Entropy-stable summation-by-parts discretization of the Euler equations on general curved elements
(Elsevier, 2017-12-17) Crean, Jared; Hicken, Jason E.; Del Rey Fernández, David C.; Zingg, David W.; Carpenter, Mark H.
We present and analyze an entropy-stable semi-discretization of the Euler equations based on high-order summation-by-parts (SBP) operators. In particular, we consider general multidimensional SBP elements, building on and generalizing previous work with tensor–product discretizations. In the absence of dissipation, we prove that the semi-discrete scheme conserves entropy; significantly, this proof of nonlinear stability does not rely on integral exactness. Furthermore, interior penalties can be incorporated into the discretization to ensure that the total (mathematical) entropy decreases monotonically, producing an entropy-stable scheme. SBP discretizations with curved elements remain accurate, conservative, and entropy stable provided the mapping Jacobian satisfies the discrete metric invariants; polynomial mappings at most one degree higher than the SBP operators automatically satisfy the metric invariants in two dimensions. In three-dimensions, we describe an elementwise optimization that leads to suitable Jacobians in the case of polynomial mappings. The properties of the semi-discrete scheme are verified and investigated using numerical experiments.
Experimental analysis of thermo-acoustic instabilities in a generic gas turbine combustor by phase-correlated PIV, chemiluminescence, and laser Raman scattering measurements
(Springer Verlag, 2015) Arndt, Christoph M.; Severin, Michael; Dem, Claudiu; Stöhr, Michael; Steinberg, Adam M.; Meier, Wolfgang
A gas turbine model combustor for partially premixed swirl flames was equipped with an optical combustion chamber and operated with CH4 and air at atmospheric pressure. The burner consisted of two concentric nozzles for separately controlled air flows and a ring of holes 12 mm upstream of the nozzle exits for fuel injection. The flame described here had a thermal power of 25 kW, a global equivalence ratio of 0.7, and exhibited thermo-acoustic instabilities at a frequency of approximately 400 Hz. The phase-dependent variations in the flame shape and relative heat release rate were determined by OH* chemiluminescence imaging; the flow velocities by stereoscopic particle image velocimetry (PIV); and the major species concentrations, mixture fraction, and temperature by laser Raman scattering. The PIV measurements showed that the flow field performed a “pumping” mode with varying inflow velocities and extent of the inner recirculation zone, triggered by the pressure variations in the combustion chamber. The flow field oscillations were accompanied by variations in the mixture fraction in the inflow region and at the flame root, which in turn were mainly caused by the variations in the CH4 concentration. The mean phase-dependent changes in the fluxes of CH4 and N2 through cross-sectional planes of the combustion chamber at different heights above the nozzle were estimated by combining the PIV and Raman data. The results revealed a periodic variation in the CH4 flux by more than 150 % in relation to the mean value, due to the combined influence of the oscillating flow velocity, density variations, and CH4 concentration. Based on the experimental results, the feedback mechanism of the thermo-acoustic pulsations could be identified as a periodic fluctuation of the equivalence ratio and fuel mass flow together with a convective delay for the transport of fuel from the fuel injector to the flame zone. The combustor and the measured data are well suited for the validation of numerical combustion simulations.
Experimental Investigation of the Thermal Properties of Tailored Expansion Lattices
(Springer Nature, 2009-05) Steeves, Craig A; Mercer, Chris; Antinucci, Emilio; He, Ming Y; Evans, Anthony G
Composite bimaterial lattice structures which possess both low, tailorable thermal expansion and nearly optimal stiffness have been proposed for applications which require high structural stiffness in environments which include large temperature fluctuations, such as the surfaces of high-speed aerospace vehicles. An experimental validation of the thermal properties of these lattices when they are constructed of practical materials with easily manufactured bonded joints is contained herein. Bonded lattices, comprising aluminum and titanium alloys, have been manufactured with press-fit dovetail joints and tested in a variety of thermal environments. Results for equilibrium heating, rapid transient heating and thermal cycling leading to shakedown are presented and shown to be consistent with theoretically and numerically attained results.
Experimental Study of Turbulence-Chemistry Interactions in Perfectly and Partially Premixed Confined Swirl Flames
(De Gruyter, 2014-12) Dem, Claudiu; Stöhr, Michael; Arndt, Christoph M.; Steinberg, Adam M.; Meier, Wolfgang
A gas turbine model combustor (Turbomeca Burner) for premixed methane/air flames has been operated at atmospheric pressure in two different modes of premixing. In the partially premixed mode, fuel was injected into the air flow within the swirl generator shortly upstream of the combustion chamber while in the perfectly premixed mode fuel and air were mixed far upstream. The main objective of this work is the study of the influence of the mode of premixing on the combustion behavior. Stereoscopic particle image velocimetry has been applied for the measurement of the flow field, OH chemiluminescence imaging for the visualization of the flame shapes and single-shot laser Raman scattering for the determination of the joint probability density functions of major species concentrations, mixture fraction and temperature. The mixing and reaction progress and effects of turbulence-chemistry interactions are characterized by scatterplots showing the correlations between different quantities. To isolate effects of mixing from combustion instabilities that were frequently observed in this combustor, operating conditions without thermo-acoustic oscillations or coherent flow structures were chosen. While the mode of premixing had no major influence on the general flame behavior characteristic differences were observed with respect to flame anchoring, the flow field in the inner recirculation zone and the CO concentration level. The results further extend the data base of previous experimental and numerical investigations with this burne
Feasibility of Metallic Structural Heat Pipes as Sharp Leading Edges for Hypersonic Vehicles
(American Society of Mechanical Engineers, 2009) Steeves, Craig A; He, Ming Y; Kaen, Scott D; Valdevit, Lorenzo; Wadley, Haydn N G; Evans, Anthony G
Hypersonic flight with hydrocarbon-fueled airbreathing propulsion requires sharp leading edges. This generates high temperatures at the leading edge surface which cannot be sustained by most materials. By integrating a planar heat pipe into the structure of the leading edge, the heat can be conducted to large flat surfaces from which it can be radiated out to the environment, significantly reducing the temperatures at the leading edge and making metals feasible materials. This paper describes a method by which the leading edge thermal boundary conditions can be ascertained from standard hypersonic correlations, and then uses these boundary conditions along with a set of analytical approximations to predict the behaviour of a planar leading edge heat pipe. The analytical predictions of the thermostructural performance are verified by finite element calculations. Given the results of the analysis, possible heat pipe fluid systems are assessed, and their applicability to the relevant conditions determined. The results indicate that the niobium alloy Cb-752, with lithium as the working fluid, is a feasible combination for Mach 6 to 8 flight with a 3 mm leading edge radius.
Formation and flame-induced suppression of the precessing vortex core in a swirl combustor: Experiments and linear stability analysis
(Elsevier, 2015-05-20) Oberleithner, Kilian; Stöhr, Michael; Seong, Ho Im; Arndt, Christoph M.; Steinberg, Adam M.
The precessing vortex core (PVC) is a coherent flow structure that is often encountered in swirling flows in gas turbine (GT) combustors. In some swirl combustors, it has been observed that a PVC is present under non-reacting conditions but disappears in the corresponding reacting cases. Since numerous studies have shown that a PVC has strong effects on the flame stabilization, it is desirable to understand the formation and suppression of PVCs in GT combustors. The present work experimentally studies the flow field in a GT model combustor at atmospheric pressure. Whereas all non-reacting conditions and detached M-shaped flames exhibit a PVC, the PVC is suppressed for attached V-shaped flames. A local linear stability analysis is then applied to the measured time-averaged velocity and density fields. For the cases where a PVC appeared in the experiment, the analysis shows a global hydrodynamic instability that manifests in a single-helical mode with its wavemaker located at the combustor inlet. The frequency of the global mode is in excellent agreement with the measured oscillation frequency and the growth rate is approximately zero, indicating the marginally stable limit-cycle. For the attached V-flame without PVC, strong radial density/temperature gradients are present at the inlet, which are shown to suppress the global instability. The interplay between the PVC and the flame is further investigated by considering a bi-stable case with intermittent transitions between V- and M-flame. The flame and flow transients are investigated experimentally via simultaneous highspeed PIV and OH-PLIF. The experiments reveal a sequence of events wherein the PVC forms prior to the transition of the flame shape. The results demonstrate the essential role of the PVC in the flame stabilization, and thereby the importance of a hydrodynamic stability analysis in the design of a swirl combustor.
High-Speed Tomographic PIV and OH PLIF Measurements in Turbulent Reactive Flows
(Springer Verlag, 2014) Coriton, Bruno; Steinberg, Adam M.; Frank, Jonathan H.
High-speed tomographic particle image velocimetry (TPIV) is demonstrated in turbulent reactive flows at acquisition rates ranging from 10 kHz to 16 kHz. The 10 kHz TPIV measurements are combined with planar laser-induced fluorescence (PLIF) imaging of OH to mark the high-temperature reaction zone of the flame. Simultaneous TPIV/OH PLIF measurements are applied to the stabilization region of a weakly-turbulent lifted DME/air jet flame (ReD = 7,600) and the mixing layer of a turbulent partially-premixed DME/air jet flame (ReD = 29,000). In the lifted jet flame, vortical structures exhibit time-dependent morphological changes and eventually dissipate as they approach the flame. In the near field of the turbulent jet flame, dynamics of localized extinction are captured as coherent structures with high compressive strain rates interact with the reaction zone and subsequently break apart. The principal axis of compressive strain has a strong preferential orientation at 45 degrees with respect to the jet axis. The 3D velocity field measurements are used to evaluate biases in 2D measurements of compressive strain rates in a turbulent jet flame. The biases in the 2D measurements primarily stem from out-of-plane orientation of the principal axis of compressive strain. Comparisons with a constant density turbulent non-reactive jet (ReD = 22,600) show that the jet flame has larger coherent structures that are confined near the reaction zone. Data from the non-reactive jet are also used to evaluate effects of noise, bias, and spatial averaging on measurements of the velocity and velocity gradients.
In-plane properties of composite laminates with through-thickness pin reinforcement
(Elsevier, 2006-05) Steeves, Craig A; Fleck, Norman A
Laminated fibre reinforced composites can be reinforced by through-thickness pins to reduce their susceptibility to delamination. However, the presence of the pins creates resin pockets and disrupts the alignment of the fibres, and may thereby lead to a degradation of the in-plane strength of the composite. Experiments and numerical simulations show that the presence of through-thickness reinforcing pins decreases the tensile strength of the composite by 27%, and the compressive strength of the composite by at least 30%. It is also shown that the pattern in which the pins are inserted has a strong influence on the compressive strength. A pin pattern is identified in order to minimise fibre alignment disruption and thereby maximise the compressive strength.
Influence of combustion on principal strain-rate transport in turbulent premixed flames
(Elsevier, 2015) Steinberg, Adam M.; Coriton, B.; Frank, J. H.
The transport of principal strain-rates (sisi) was experimentally investigated using high-repetition-rate (10 kHz) tomographic particle image velocimetry (T-PIV) and OH planar laser induced fluorescence (PLIF) in a Rej=13,000Rej=13,000 turbulent premixed flame. These measurements allowed calculation of the source terms in the sisi transport equation associated with the strain-rate and vorticity fields. Furthermore, the Lagrangian derivatives of sisi could be calculated by tracking theoretical Lagrangian fluid particles through space and time using the T-PIV data. These Lagrangian derivatives and the resolved source terms allowed the combined effects of the unresolved source terms to be inferred, namely the pressure Hessian, viscous dissipation, density gradients, and viscosity gradients. Statistics conditioned on the location of the Lagrangian fluid particles relative to the flame showed slight reductions in the strain-rate and vorticity source terms in the flame, indicating that these aspects of the turbulence were attenuated by the flame. Comparing the difference between the inferred source terms in the vicinity of the flame to the non-reacting flow showed that attenuation of sisi arose due to the combined effects of density and pressure gradients in the flame. The effects of flame-induced dilatation were small relative to the turbulent strain-rate and no change was found in the relative alignment of vorticity and strain-rate in the flame.
Learning-based Nonlinear Model Predictive Control to Improve Vision-based Mobile Robot Path Tracking
(Wiley, 2015-06-15) Ostafew, Chris J.; Schoellig, Angela P.; Barfoot, Timothy D.; Collier, Jack
This paper presents a Learning-based Nonlinear Model Predictive Control (LB-NMPC) algorithm to achieve high-performance path tracking in challenging off-road terrain through learning. The LB-NMPC algorithm uses a simple a priori vehicle model and a learned disturbance model. Disturbances are modeled as a Gaussian process (GP) as a function of system state, input, and other relevant variables. The GP is updated based on experience collected during previous trials. Localization for the controller is provided by an onboard, vision-based mapping and navigation system enabling operation in large-scale, GPS-denied environments. The paper presents experimental results including over 3 km of travel by three significantly different robot platforms with masses ranging from 50 to 600 kg and at speeds ranging from 0.35 to 1.2 m/s (associated video at http://tiny.cc/RoverLearnsDisturbances). Planned speeds are generated by a novel experience-based speed scheduler that balances overall travel time, path-tracking errors, and localization reliability. The results show that the controller can start from a generic a priori vehicle model and subsequently learn to reduce vehicle- and trajectory-specific path-tracking errors based on experience.

Browse

Browsing University of Toronto Institute for Aerospace Studies (UTIAS) by Title