Verification and validation of an active sailing simulator

Transcription

1 Verification and validation of an active sailing simulator Jonathan R. Binns; Nicholas Clark, Troy Munro Australian Maritime College, University of Tasmania Chris Manzie; Denny Oetomo; Graham Bennett University of Melbourne Norman Saunders; Mark Habgood University of Melbourne and Virtual Sailing, Pty Ltd Abstract. Virtual Sailing Pty Ltd is an Australian company that for over 10 years has invested significant time and resources into developing a ride on sailing simulator, which replicates several classes including the Laser and 29er sailing dinghies. The simulator provides a means for training and performance assessment of sailors, including the feature of providing rehabilitation and training for disabled body sailors through the V-Sail Access Simulator. Previously much effort has been devoted to system identification of larger marine vessels; this is not the case with small sailing craft that are being replicated by the simulator. The need for a validated simulation model, combined with reducing costs of computing technology, has led to the creation of a wireless manoeuvring measurement network. The developed network has been shown to be of great use to the difficult scenario of a small sailing vessel operating in a harsh environment. This has a potential to be used on many different platforms. A series of manoeuvring tests was conducted in order to obtain a Laser class sailing dinghy manoeuvring model for Verification, Validation and Calibration (VVC) of the current simulation model. The manoeuvring model has been compared to tests by repeating the full scale manoeuvres identified. 1. INTRODUCTION Over the past 10 years, Virtual Sailing Pty. Ltd. has continuously developed a human in the loop (HITL) simulator, dubbed the VSail Trainer. The trainer is able to simulate a number of classes, including: Laser Standard; Laser Radial; Laser 4.7; Optimist; Byte; Mega Byte; Liberty, and 29 er. Applications of the VSail Trainer [9] are wide and varied, including: Training of beginners through to Olympic level sailors (Figure 1); The assessment of tactical and physical performance of elite level sailors, in a controlled environment, with data logging and simulation replay facilities; Training of disabled sailors, and Rehabilitation of spinal injury patients, utilising the rolling motion of the simulator. In line with continual development of the VSail Trainer, the team of researchers working on the VSail Trainer aims to validate the current simulation to ensure high fidelity to real life. Figure 1: A VSail Trainer suite, including dynamic rolling hull and on screen virtual course position. Source: The Virtual Sailing Simulator has been continuously analysed and developed [2, 4, 13] over the last ten years. The simulated yacht system is based on a Velocity Prediction Program (VPP) [11], with sail lift and drag coefficients derived from [7] and modified to suit the simulated sail parameters. This development approach has led to a sailing simulator with performance that feels like on the water. At this stage in the life of the simulator and given that an elite sailor/ discerning customer may request hard evidence to show that the simulator bears fidelity to real lift, a validation study of the simulator was required Simulation Verification, Validation and Calibration (VVC) is common practice [8], [10] during the development/upgrade life-cycle phase of a simulation program. Specifically, the ITTC outlines procedures and guidelines [6] that are to be followed when validating a maritime vehicle/simulation. However, the guidelines have been developed to suit marine vehicles with engine propulsion, not sails, leading to the need for a more suitable series of manoeuvre tests. [3] outlines a modified series of manoeuvres that are designed specifically to obtain manoeuvring coefficients for sailing yachts. The authors have developed a low cost wireless Data Acquisition System (DAS) to carry out yacht system, or characteristic dynamic movement identification (not limited to manoeuvring coefficients) [1]. To suite the resources offered by the DAS, select manoeuvres outlined by [3] such as luffing into the wind, and monitoring rudder angle during tacking angles, while logging position/velocity/heading with GPS were used. Additional resistance validation data for a full scale laser dinghy has been published by [5], and it is recommended that a comparison be made with the simulation drag curve in future.

2 Once raw test data has been obtained, postprocessing of all data is required. In order to obtain dinghy heading, True Wind Angle (TWA) and True Wind Speed (TWS) relative to the course, Longitude and Latitude information needs to be converted to 360 degree heading by using a modified version [12] of Great Circle Navigation Formulae [14]. With time series of velocity and heading obtained, assessment of yacht performance is commonly represented by a polar plot of velocity over land through 360 heading with assumed constant wind direction[2]. 2. EXPERIMENTAL SETUP PROCEDURES The following is an introduction to key experimental setup procedures and a brief description of the equipment used. GoPro HD cameras were placed on the forward deck (Figure 2), facing forward and aft, to give video footage of course direction, pilot position and rudder position. Video footage of the dinghy leaving and arriving at the start/finish pontoon combined with video timestamps, formed a part of time synchronising the GPS and onboard data channels. Figure 4: Rudder encoder belt attachment. Rudder calibration took place using markings placed on the aft deck with a degrees wheel (Figure 5), so that rudder angle could be obtained from the raw encoder information. At the start of each test session, the rudder was held centred then stepped through: 20 degree steps every ~5 seconds to starboard, centre, port and centre, and 10 degrees every 5 seconds to starboard, centre, port and centre. Figure 2: Dark (Micro-computer) and Light grey (Transmitter) boxes attached to the deck. The Port/Aft facing camera is attached, with a plate ready for starboard camera mounting. The mast encoder can be seen to the central left, whilst the base of the anemometer is to the extreme right at the bow. Figure 5: Rudder angle markings drawn on the aft deck. Deck markings were used before each test to obtain rudder angle calibration data. The rudder was rotated through incremental steps of 10 and 20 from side to side to a maximum of 60, starting at the centre. 2.2 Boom angle attachment and calibration The mast encoder designed to measure boom rotation angle was tied to the mast with a piece of elastic cord. Recording boom angle through mast rotation is possible since boom angle relative to the mast is fixed at boom/mast joint. Figure 3: View from onboard camera, used to identify manoeuvres. 2.1 Rudder angle attachment and calibration A rudder encoder was attached to the rudderstock using an elastic cord (Figure 4). Figure 6: Mast Encoder Attached with three loops of fine elastic cord around both mast and encoder.

3 Using a similar calibration procedure to rudder angle calibration, the boom was held in line with the centreline at 0 rotation, then stepped through 45, 90 and 180 rotation angles through both the port and starboard sides. Figure 7: DAQ Laptop, and GPS unit (bottom right of bag) in waterproof bag. Bag is shackled to the deck using plates attached to the deck and d- shackles. 2.3 Data storage The onboard DAS used a sealed notebook in a waterproof bag to collect recorded data. The bag was anchored to the deck with d shackles (Figure 7). A Velocitek SC-1 GPS unit was located inside the cockpit of the dinghy. It was housed in a waterproof case specifically designed for use in the marine environment. 3. ON-WATER TEST MANOEUVRES The following manoeuvres were conducted on Albert Park Lake, Melbourne, with the aim of obtaining full scale validation data in accordance with the aims of this study. 3.1 Test 1 Complete laps of a set course The marker buoys with anchor weights were dropped in the lake to form a triangular course (Figure 8). Multiple laps were completed by rounding the course to the port side; starboard side and upwind/downwind hotdog laps. All position data were recorded by GPS. 3.2 Test 2 Full tacking manoeuvre Tacking (changing direction when sailing upwind) performance of sailor, yacht and simulator is an area that when improved can reduce course times substantially. Tack manoeuvres required logging: so in order to sail to the upwind course marks, multiple tacks were carried. In addition to tacks completed during lap runs, multiple consecutive upwind tacks were also completed. Figure 8: Comparison of recorded track overlay (top) and Post processed track (bottom). 3.3 Test 3 Stopping close hauled by luffing into the wind This manoeuvre was carried out by sailing upwind at a stable speed, then turning the dinghy into the wind so that the sail experienced zero angle of attack. This test was conducted to obtain velocity data that could be used to examine the combined calm water hydrodynamic resistance of the hull, underwater appendages and aerodynamic drag of the sails. The data have not yet been used, but are available for future work. 4. SAILING SIMULATOR TESTING AND DATA ACQUISITION The following section discusses the procedure, analysis conducted, results and conclusions obtained in carrying out validation of the existing sailing simulation based on the newly acquired full scale test data outlined in Section 3. Figure 9: Dynamic sailing simulator input was used to realistically replicate real world conditions. 4.1 Introduction Testing on the sailing simulator is intended to replicate real world sailing as closely as possible, so that

4 validation of data remains relevant. The same environmental parameters including wind speed, course orientation, and course size were programmed into the simulator. With conditions set, the same manoeuvres that were completed at Albert Park; e.g. lapping a set course and completing consecutive upwind tacks were carried out. 4.2 Experimental setup Markers were added to the simulation, using GPS data and wind direction information obtained from Albert Park Testing. Wind conditions were set to 10 knots, with no gusts. The simulator was set to its dynamic mode that required the pilot s bodyweight to provide a righting moment. 5. COMPARISON OF SIMULATOR TO REAL SAILING 5.1 Lap times The circuits completed, shown in Figure 8 were typically sailed in times within 10% variation between the actual on-water performance compared to the simulated environment. Overall this was very encouraging and provided the first real comparison of this simulator to on-water performance. 5.2 Rudder feedback In a real sailing vessel the skipper is able to feel the amount of force required to steer the vessel through the tiller (the connection between the skipper and the rudder) from the hydrodynamic forces on the rudder. This feedback force adds a degree of complexity to the simulation; although not a large undertaking this does add cost and potential maintenance issues. Therefore the need for such an enhancement should be carefully assessed. The assessment was first carried out qualitatively by asking the user of the simulator, who was also a sailor with on-water experience, how realistic the rudder feedback felt in terms of the magnitude of the force and the response times of the force. Secondly, the actual rudder angles were compared for the simulated and real cases. To gain a baseline for the study, the rudder angles were also compared for the same user with the feedback turned off. The results of the qualitative answers on the rudder feedback have been replicated in Table 1. Although only one user has been sampled here, it does point the way forward in terms of how to improve the system Quantitative results of actual rudder angles measured for the passive rudder system, the active force feedback system and the actual on-water performance are plotted in Figure 10. From this data, and many other tacking manoeuvres analysed, it was immediately apparent that the on-water data were giving a larger magnitude of rudder response with the shortest time. The passive system closely follows that of the on-water data. However the passive system requires more time to tack the vessel. The active system gave the least similar response as the rudder angles continue to fluctuate after the tack was performed and required a greater amount of time get into a constant heading. Combining this with the comments detailed above, it appears that the given the limited data collected; the active rudder feedback has produced something that feels better for the user once settled, but which certainly needs quicker response times to induce realistic rudder responses. Table 1 Torque Feedback Results Questions Response from Ranging from 0-10, how realistic to on the water sailing are the forces and response time of the new torque feedback? Does the force seem excessive or not strong enough? In regards to the rudder response, are the changes in force jumpy or was there a smooth transition between the varying forces? Do you think the effects of the rudder torque feedback have increased the realism of the simulator? 6. CONCLUSIONS operator 7/10. It was quite realistic, however takes about 2-3 seconds for the force to build up to what I expected. Upwind tacks feel spot on with regards to force and direction. No, I found it very smooth especially when I m tacking and gybing. It definitely gave a better realistic feel when heading upwind, as it pulls harder in my hand when the wind becomes stronger. A wireless system has been developed to measure multiple degrees of freedom of a sailing dinghy. This system has been successfully used to measure on-water performance of a Laser class sailing dinghy. A series of manoeuvring tests were performed and repeated within the simulated environment. Overall sailing times compared well between actual onwater measurements and those recorded in the simulator. In addition, an active rudder force feedback unit was tested. Although qualitative assessment of the force feedback from the sailor might suggest that such a unit increased realism, the rudder data collected indicated problems with the force feedback with a clear indication on the area that requires improvement. Future research will focus on increasing the number of experienced sailors used for assessment and on improving the response times of force feedback systems in the rudder. REFERENCES 1. Bennett, G., Manzie, C., Oetomo, D., Binns, J.R., & Saunders, N.R., A wireless sensor network for

5 Rudder Angle ( ) system identification of sailboat dinghies., in Simtect : Brisbane, Australia. 2. Binns, J.R., Bethwaite, F.W., & Saunders, N.R., (2002), Development of a more realistic sailing simulator. in The 1st High Performance Yacht Design Conference. Auckland. 3. Binns, J.R., Hochkirch, K., De Bord, F., & Burns, I.A., (2008), The development and use of sailing simulation for IACC starting manoeuvre training. in 3rd High Performance Yacht Design Conference. Auckland, New Zealand. 4. Binns, J.R., Maher, R., Chin, K.H., & Bose, N., (2009), Development and use of a computer controlled sailing simulation. in SimTect. Adelaide, Australia: Simulation Industry Association of Australia. 5. Carrico, T., (2005), A Velocity Prediction Program for a Planing Dinghy. in The 17th Chesapeake Sailing Yacht Symposium. Annapolis, Maryland. 6. ITTC, M.C.o.r., Validation of manoeurvring simulation models., ITTC, Editor Marchaj, C.A., (1988), Aero-Hydrodynamics of Sailing. Adlard Coles Nautical: London, UK,. p McFarlane, D. & Office, A.D.S., Simulation verification, validation and accreditation guide., Department of Defence, Editor. 2002, Commonwealth of Australia: Canberra, ACT. 9. Mooney, J., Saunders, N.R., Habgood, M., & Binns, J.R., (2009), Multiple applications of sailing simulation. in SimTect. Adelaide, Australia: Simulation Industry Association of Australia. 10. Rakha, H., Hellinga, B., Van Aerde, M., & Perez, W., (1996), "Systemic verification, validation and calibration of traffic simulation models.". 11. Richardt, T., Harries, S., & Hochkirch, K., Maneuvering simulations for ships and sailing yachts using FRIENDSHIP-equilibrium as a open modular workbench., in International Euro- Conference on Computer Applications and Information Technology in the Maritime Industries. 2005: Hamburg, Germany. 12. Rick, D., (2001), Bearing Between Two Points [cited October 2011]; Earth Quadrant specific Greater Circle Bearing Formulae]. Available from: ml. 13. Walls, J.T., Gale, T., Saunders, N.R., & Bertrand, L., (1998), "Assessment of upwind dinghy sailing performance using a virtual reality dinghy sailing simulator." The Australian Journal of Science and Medicine in Sport, 1: p Williams, E., (2011), Aviation Formulary V1.46. [cited October 2011]; Greater Circle Navigation Formulae]. Available from: Passive Active On-Water Time (s) Figure 10: Comparison of full scale on water rudder angle during tack manoeuvres, compared to simulated tacks and simulated tacks with active rudder feedback. Initial results show that maximum rudder angle during manoeuvres in both full scale and simulated sailing to be ~40 degrees, showing that the simulator closely replicates real life.