26 TH INTERNATIONAL CONGRE OF THE AERONAUTICAL CIENCE ANALYI OF DEEPTALL LANDING FOR UAV Hiroki Taniguci* *Te University of Tokyo Keywords: stall landing, UAV, landing metod Abstract Deepstall landing is one of te landing metods by wic airplanes can decrease fligt speed wile maintaining a deep glide pat angle. Tis is because te wing can generate large air drag in te deep stall state. Terefore, te metod is suitable for small unmanned aerial veicle to land witin narrow areas surrounded by tall obstacles. Tis paper considers te trim condition in te deepstall fligt pase and a numerical fligt simulation from te cruise to te deepstall landing. Furtermore, fligt tests were conducted to confirm te deepstall landing caracteristics. 1 Introduction A small unmanned aerial veicle (UAV) for aerial potograpy and remote sensing as been developed. In general, te size of a small UAV can range from several tens of centimeters to a few meters, and te propulsion is mostly by an electric powered motor. Ground control software can run on a laptop computer, so te ground station can be set up practically everywere. Generally, small UAVs are deployed in environments were no access to conventional runways is available, because te UAV can take off by and launc or catapult, and land at a sort distance. Moreover, it is often andled at confined areas surrounded by trees or buildings. It is very difficult to land UAVs at tese areas adopting normal glide landing metods, especially for untrained operators. Hence, some effective landing metods must be devised so tat te small UAV system can develop into a user-friendly tecnology. In tese environments, te orizontal speed needs to be decreased and te deep pat angle as to be maintained to land safely. To meet tis requirement, several solutions ave been considered, e.g., paracute landing and vertical take off and landing (VTOL) mecanism. Te paracute landing is te most common metod because it is easy to operate and te UAV can be retrieved at any time and anywere. However, it requires a considerable skill to properly fold paracutes and store tem in te aircraft body. Furtermore, te paracute itself as large air drag, wic makes it difficult to precisely land te UAV at te target point in strong wind conditions. On te oter and, adopting te VTOL metod, we can andle UAVs in narrow spaces. However, payload and endurance are limited and ig performance controller is necessary as te VTOL system is complicated and originally unstable. Anoter way wic doesn t suffer from tese disadvantages is deepstall landing, wic is studied in tis paper. Te mecanism is quite simple and it is an effective way to land te planes in narrow space. Because of te large air drag at deep stall, te orizontal speed can be reduced effectively wile te deep pat angle is kept. ome landing metods taking advantage of stall ave been reported in literature [1,2], but deepstall landing metod was originally employed to te free fligt model plane and as been developed empirically. o, te caracteristics of deepstall landing ave not been studied precisely. In tis paper, trim state at deepstall landing and te performance as landing metod are considered using trim state contour graps. Numerical fligt simulation takes into account te istory of attitude and speed from cruise to 1
Hiroki Taniguci deepstall pase. A model plane wit measuring instruments was developed and a fligt test was done to confirm computational results. 2 Deepstall Landing Metod 2.1 Caracteristics of deep stall Deep stall is te state in wic te angle of attack is muc iger tan te stall angle. Even toug te flow is unstable around te stall angle, it is stable in post-stall. Terefore, te wing can generate constant lift and drag forces wen te angle of attack exceeds te stall angle by a certain value. Fig.1 and Fig.2 sow te wind tunnel data of a NACA0012 wing [3]. According to Fig.1, te lift coefficient C L and drag coefficient C D are larger tan at cruise. And Fig.2 means tat te value of te lift to drag ratio L/D is smaller in te post stall state tan in normal fligt. Fig. 1 C L and C D data of NACA0012 te stall angle. Ten te plane trims at te deep stall state because te lift of te HTP cancels te wing and body moments. Furtermore, lift of te HTP acts as a restoring force because te HTP is not in stall. Consequently, te plane can keep tis stable trim state. In tis trimmed deep stall state, C L and CD are so large tat fligt speed is decreased effectively. Besides, te L/D is so small tat te plane can keep te deep glide pat angle. Fig. 3 Model of deepstall landing 2.3 Advantages of Deepstall Landing Te deepstall landing metod can be implemented wit a simple mecanism of HTP and it as a good stability. Terefore, untrained operators can easily andle it. Tis is an important point because small UAV system is becoming a platform for aerial potograpy and remote sensing and in many cases non-skilled operators andle te system. Furtermore, te pat angle and speed can be set arbitrarily, controlling te HTP to tilt up by an angle δ. Hence, te landing point can be adjusted in strong wind conditions. On te oter and, because touc down speed is iger tan for te conventional metods, its application may be limited to small UAV only. Fig. 2 L/D of NACA0012 2.2 Deepstall Landing Mecanism Te only mecanism to get te plane into deep stall is quickly tilt of te orizontal tail plane (HTP). Wen te HTP tilts up suddenly at cruise, te plane pitces up largely and te angle of attack gets muc iger tan 2
ANALYI OF DEEPTALL LANDING FOR UAV Fig. 4 Caracteristics of landing metods for small UAV system 3 Trim tate Analysis 3.1 Equations of Equilibrium Firstly, longitudinal trim conditions in te pase of deepstall landing were estimated. Te trim state of an aircraft can be expressed by te following tree equations: L cosγ + Dsin γ = M g (1) Lsinγ Dcosγ = 0 (2) M L l cos( α + δ) = 0 (3) Variables are defined in Fig.3. Constant number g is a gravitation constant. Variables wit suffix are te value for te HTP. Lift and drag forces and moment can be modeled as functions of te angle of attack referring to NACA0012 wind tunnel data wic includes after stall data. Te moment balance equation (3) is expressed as below applying HTP volume V as variable. C V C cos( α + δ) = 0 (4) M L l V = (5) c Terefore, te trim state can be determined once V and δ are fixed. 3.2 Calculation Conditions Te plane model is based on a existing free fligt plane, wic weigs 230g, and wing span is 1.4m long. Te plane as a pop up mecanism of te HTP and deepstall landing as been realized. Its details are summarized in Table.1. In tis analysis, trim state was calculated under te condition tat V and δ were te variable and weigt M and wing area were constant. Table. 1 Configuration of te model free fligt plane General Main wing I yy M b AR weigt inertia wing span aspect (kg) moment area (kg m 2 (m) ) (m 2 ratio ) 0.231 0.00622 1.38 0.159 12 HTP b l AR V wing span 1 aspect HTP area (m) (m 2 ) (m) ratio volume 0.337 0.0256 0.635 4.43 1.12 1 expressed in Fig.3 3.3 Contour Lines of Trim tate Fig.5 to 7 sow te contours of te attitude and speed at trimmed deepstall landing pase. Horizontal and vertical axes of te graps sow V and δ respectively. Fig.5 sows te pitc angle contour lines at te final deepstall state. Te red lines mean tat pitc angle is 0 and te plane can land orizontally. However, te pitc angle is less tan 6 in most cases, so te plane can land almost orizontally in any configurations. Final velocity decreases as V and δ increase. Tis appens because te angle of attack becomes larger at iger V and δ. Fig.7 sows te final vertical speed and pat angle contours. Tese are important parameters to measure te performance of deepstall landing because vertical velocity directly defines te impact to te body and pat angle decides te size of area necessary to land. Te deepstall 3
Hiroki Taniguci constraint line was added because te angle of attack is smaller tan te stall angle below te line. As tilt up angle δ increases, vertical speed becomes larger but te deep pat angle can be maintained. Terefore, an optimal configuration must be cosen depending on te landing area environments and damage to te body. and pat angle istories wen te tilt up angle δ is 45. It takes about 2.5 seconds to converge to te final trim state, wic means te attitude of te plane can stabilize quickly. Fig. 5 Pitc angle Fig. 6 final absolute speed Fig. 8 Trajectories and attitudes of te deepstall simulation Fig. 7 Performance curve of deepstall landing 4 Fligt imulation In te next step, transition fligt from te cruise into te deepstall landing was simulated in a fligt simulator developed in MATLAB/imulink. It considers te fligt in te longitudinal plane only. As in te trim analysis, lift and drag forces and moment are modeled using NACA0012 wind tunnel data. Fig.8 sows te resulting trajectories of te airplane wen te HTP is tilted up 25,35,45 and 55 respectively. Plane figures plotted every 0.5 second sow te position and attitude of te plane. Te HTP is tilted up after 1.0 second and simulations are carried out until 10 second. As described in section 2, te glide pat angle is larger at iger tilt up angle. Tese results indicate tat first te airplane pitces up, ten it pitces down and it comes to be nearly orizontal in te end. Fig.9 sows te speed and te angle of attack, pitc Fig. 9 peed and angle istories at 45 tilted up 5 Analysis Result of Real Fligt Tests Fligt tests were conducted to confirm te caracteristics of deepstall landing in real fligt. Fligt data was obtained and it was compared wit te simulation results. 5.1 Test plane Te sape of te test plane is similar to te model free fligt plane used in te computational simulation. Te configuration of te plane is summarized in Table 2. It is a motor 4
ANALYI OF DEEPTALL LANDING FOR UAV powered glider, wit te HTP fixed to te body by a ig precision servomotor. Terefore, te wing can be tilted up arbitrarily from 5 to 50 by pilot commands. A fligt logger, wic was composed of 3 gyros, 3 accelerometers and a GP receiver, was fixed in te plane. It ad been developed in te laboratory and te accuracy ad been confirmed [4]. As a result, attitude, speed, and position istory data were recorded. Fig. 10 Test plane 5.2 Fligt Test data Using tis test plane, a fligt test was conducted and data could be obtained successfully. Fig.12 sows te 3D trajectory and plane attitude of fligt test. Te HTP tilt up angle δ was 35. Te plane direction was rotating as te roll angle was not equal to zero, because te HTP was accidentally fixed wit a little tilt. Fig.13 sows te trajectory and attitude projected on te longitudinal plane. Wen te HTP was tilted up, te plane pitced up and ten down, and it converged to te orizontal trim state. Fig.14 sows te speed and attitude time istories. Blue lines are fligt test data and red ones are simulation results. Te same trend is observed in te fligt test data and te simulation results discussed in section 4. Te value of te glide pat angle is a little different between te anterior alf and posterior alf of te deepstall pase. Tis could be because te wind direction relative to te plane canged because te plane was rotating. Fig. 11 Fligt logger installed in te test plane Fig. 12 3D plot of fligt trajectory and te plane attitudes Table. 2 Configuration of te test plane General Main wing M I yy b AR (kg) (kg m 2 ) (m) (m 2 ) 0.498 0.0227 1.52 0.190 12 HTP b l AR V (m) (m 2 ) (m) 0.37 0.0315 0.603 4.35 0.828 Fig. 13 Trajectory and te plane attitude projected on longitudinal plane 5
Hiroki Taniguci [3] Critzos C, Heyson H, Boswinkle, Aerodynamic caracteristics of NACA0012 airfoil section at angles of attack from 0 to 180, NACA TECHNICAL NOTE 3361, 1955. [4] Naruoka M. Attitude accuracy improvement of ultra low-grade MEM IN using alignment of GP antenna. 26t Congress of International Council of Aeronautical ciences (ICA 2008), Ancorage, 2008. Fig. 14 peed and angle istories of real fligt test and simulation result 6 Conclusions Adopting deepstall landing metod, te plane can decrease te fligt speed effectively and maintain te deep pat angle. Tese advantages were confirmed troug trim analysis. Using te trim plot (Fig.7), te values of V and δ can be decided for a certain pat angle, steep enoug to land in a confined area and wic result in a vertical speed low enoug to prevent damage of te body. Te longitudinal transition pase from steady fligt to deepstall landing was simulated and an adequate dynamical stability was observed. Finally, a real fligt test was conducted and te fligt data support tat te trim analysis and fligt simulation are effective to estimate te dynamical caracteristics and performance of deepstall landing. In a future study, more researc about longitudinal and lateral stability will be done and UAV design tat is optimal for deepstall landing will be considered. Copyrigt tatement Te autors confirm tat tey, and/or teir company or institution, old copyrigt on all of te original material included in teir paper. Tey also confirm tey ave obtained permission, from te copyrigt older of any tird party material included in teir paper, to publis it as part of teir paper. Te autors grant full permission for te publication and distribution of teir paper as part of te ICA2008 proceedings or as individual off-prints from te proceedings. References [1] Crowter W. Perced Landing and Takeoff for Fixed Wing UAVs. NATO AVT ymposium on Unmanned Veicles for Aerial, Ground and Naval Military Operations, 2000 (Report #RTO-MP-052). [2] Crowter W, Prassas K. Post-stall landing for field retrieval of unmanned air veicles. 14t Bristol International unmanned air veicle systems conference, April 1999 (Paper #39). 6