A Low-Cost GPS-Based Wave Height and Direction Sensor for Marine Safety Masatoshi Harigae, Isao Yamaguchi, Tokio Kasai and Hirotaka Igawa, Japan Aerospace Exploration Agency (JAXA) Hiroto Nakanishi and Takahiro Murayama, Japan Weather Association (JWA) Yasunori Iwanaka, Zeni Lite Buoy Co., Ltd. Hirotaka Suko, Furuno Electric Co., Ltd. Hiroto Nakanishi is a technical planning manager at the Japan Weather Association. He has about 15years experience of air pollution data and weather data collection systems. BIOGRAPHY Masatoshi Harigae is the director at the Advanced Control Research Group of the Institute of Space Technology and Aeronautics of JAXA, and has over 15 years experience in the development of GPS/INS hybrid navigation systems and their application to the air traffic system. He is also engaged in GPS technology transfer to commercial fields. He received B.S., M.S. and Ph.D. degrees in aerospace engineering from the University of Tokyo. Takahiro Murayama is an engineer at the Applied Meteorology Research Section, Research Department of JWA. He is involved in the development of a sea wave observation system with buoys and the atmospheric and sea wave numerical simulations. He received a B.E. degree in Marine Science and Technology from Tokai University. Isao Yamaguchi is a senior researcher at the Structure Research Group of the Institute of Space Technology and Aeronautics of JAXA. His research interests are dynamics, system identification, modeling and control of spacecraft and large space structures. He received B.S., M.S. and the Ph.D. degrees in aeronautical engineering in 1981, 1983 and 1997, respectively from the University of Tokyo. He is a member of SICE, JSME, The Japan Society for Aeronautical and Space Sciences, AIAA and IEEE. Yasunori Iwanaka is a chief of technical development department of Zeni Lite Buoy Co., Ltd. He has about 5 years experience of marine observation monitoring systems. He has a B.S. in electronics engineering from Osaka Institute of Technology. Hirotaka Suko is a software engineer at Furuno Electric Co., Ltd., and has over 10 years experience in the development of GPS receivers for navigation, time synchronization and terrain remote monitoring. He has a B.S. in management engineering from Setsunan University. Tokio Kasai is a senior researcher at the Structure Research Group of the Institute of Space Technology and Aeronautics of JAXA. His research interests include structural dynamics, system identification and modeling of flexible spacecraft. He received B.S. and M.S. degrees in mechanical engineering from Waseda University. ABSTRACT Predicting the development, decay and propagation of ocean waves is essential to allow maritime operations to be carried out safely and economically. A low cost wave height and direction sensor that can be deployed in the open sea is expected to improve the accuracy of wave prediction. We propose a new method for measuring wave height and direction by installing a point-positioning GPS receiver on a buoy placed in the open sea. The essential idea of measuring wave height with centimeter accuracy depends on the fact that the most of the spectrum of GPS Hirotaka Igawa is a researcher at the Structure Research Group of the Institute of Space Technology and Aeronautics of JAXA. His research interests are structural dynamics, modeling of lightweight structures and smart structures. He received B.S., M.S. and Ph.D. degrees in mechanical engineering science in 1992, 1994 and 1997 from Tokyo Institute of Technology. ION GNSS 17th International Technical Meeting of the Satellite Division, 21-24 Sept. 2004, Long Beach, CA 1234 for waves that are several meters high. The advantages of this sensor are that there are no restrictions as to its location and it is easy to handle. It is therefore usually deployed on drifting buoys. The sensor cost, especially for 3-axis accelerometers, is generally high. point-positioning error exists in the frequency domain below 0.01 Hz, but the frequency spectrum of ocean waves is in a separate band at around 0.1 Hz. Therefore, a sophisticated high-pass filter can extract the movements of a GPS equipped buoy excited by ocean waves with minimum effect of GPS point-positioning error. The three-dimensional GPS positioning data also enables us to measure wave direction. Our proposed system has been already adopted by the Japan Meteorological Agency as the world’s first GPS-based wave height and direction measuring system. This paper presents the concept of the point-positioning GPS-based wave sensor, the algorithms for filtering and extracting wave data, and the results of field trials carried out in the open sea. Another approach to measuring wave height and direction is based on kinematic GPS. Kinematic GPS can achieve centimeter-level positioning of a floating buoy in horizontal and vertical directions and measure wave height and direction. Kinematic GPS has the advantage of being able to measure long-period waves with small accelerations, such as tsunami and tidal waves. However the operational range of kinematic GPS is presently limited to about 20 km from the shore even using dual-frequency GPS receivers, and the system is complex and costs are very high. INTRODUCTION The current prediction of ocean waves based on CFD (Computational Fluid Dynamics) does not satisfy the requirements of maritime navigation because of the poor accuracy of ocean wave models. A low cost wave height and direction sensor that can be deployed in the open sea is expected to provide data that will allow the prediction models to be improved. We propose a novel method that uses only a low-cost point-positioning GPS receiver (consumer car navigator class is acceptable) installed on a buoy that extracts wave height with an accuracy of several centimeters and direction with an accuracy of 5 degrees. This method does not require a pair of GPS receivers, and so can be deployed even in the middle of the Pacific Ocean. The essential idea of measuring wave height with centimeter accuracy depends on the GPS measurement characteristic that the most of the frequency spectrum of GPS pointpositioning error exists below 0.01 Hz, while the frequency spectrum of ocean waves is in a separate band at around 0.1 Hz. Therefore, a sophisticated high-pass filter can extract the movement of a GPS-equipped buoy excited by ocean waves with minimum influence of GPS point-positioning errors (U.S. Patent Application No. P7153-2069-030397). While there are several ways to measure wave height and direction, ultrasonic sensors are widely used as the most reliable sensors, and Japan has constructed an observation network of such sensors. Ultrasonic sensors measure wave height with a design resolution of several centimeters by measuring the distance to the sea surface from an observation device installed on the seabed by the emission of ultrasonic waves. There is also a wave sensor with canted ultrasonic beams that measures wave direction. However, the measurable distance from the sea bed to the sea surface is limited to around 50 m, so the sensors cannot be deployed in the open ocean but only in littoral areas. It should be also noted that initial installation and maintenance costs are usually very high. This new GPS-based system does not require a complicated algorithm to extract wave data, but simply applies a high-pass filter to point-positioning data. Therefore, the algorithm can execute in the navigation processing core of a GPS receiver. We have developed a GPS receiver board with a wave height and direction measurement function. This transmits only analyzed ocean wave data to a land station via communication satellite, rather than raw GPS position data that must be processed separately. This reduces the volume of data that has to be communicated, and means our method also has low operating costs as well as low acquisition cost.. An accelerometer installed on a floating buoy can measure wave height by detecting the vertical motions of the buoy. It is easy to expand its function to measure wave direction by using 3-axis accelerometers directionally stabilized by gyros or an oil pod. Horizontal and vertical displacements can be obtained by double integration of three acceleration signal components. The vertical component of displacement gives wave height, and correlation analysis between the horizontal and vertical components gives wave direction. Measurement accuracy is 2-10% of wave height, which is several ten centimeters ION GNSS 17th International Technical Meeting of the Satellite Division, 21-24 Sept. 2004, Long Beach, CA Our proposed system has been already adopted by the Japan Meteorological Agency as the world’s first 1235 GPS-based wave height and direction measuring system. The system provides valuable ocean wave data automatically via the Internet for 24 hours continuously. It is also planned to retrofit our system to add a wave height and direction measurement capability to existing buoys. z x This paper presents the concept of the point-positioning GPS-based wave sensor, the algorithms of filtering and extracting wave data, and the results of experiments carried out in the open sea. WAVE FUNDAMENTALS 1) Movement of the ocean surface is classified as various kinds of waves according to time and space scales. Ocean waves, which we treat here, consist of wind waves and swells. The first step of a wind wave is called a capillary wave and its restitution force is surface tension. Capillary waves develop into ripples with 0.3 sec period by the effect of gravity. Ripples develop further with the aid of wind forces, and their periods reach 10-15 sec and they can sometimes grow to 10 m in height. Wind waves that develop further and move away from their origin are called swells. Swells are divided into three categories according to wavelength and period: short, middle and long swells. Short swells refer to those with a wavelength below 100 m and a period less than 8 sec. Middle swells have wavelengths 100-200 m and periods 8-12 sec. Long swells have wavelengths over 200m, sometimes reaching 400 m, with periods of up to 20 sec. Among the movements of the ocean surface, ocean waves, whose restitution force is gravity, have the most energy. It is therefore important for maritime safety to predict these gravitational waves. The motion of gravitational waves is described as follows: cosh k ( z 0 + h) x = x0 + a cos(kx 0 − ωt ), sinh kh y = y0 , z = z0 + a Fig. 1 Trajectory of water particle by gravitational wave floating buoy Fig. 2 Trajectory of water particle by surface wave cosh k ( z 0 + h) , sinh kh sinh k ( z 0 + h) Az ( z 0 ) = a . sinh kh Ax ( z 0 ) = a (1) sinh k ( z 0 + h) sin( kx 0 − ωt ) sinh kh Fig. 1 shows the motion of gravitational waves according to equation (2). Because Ax > Az > 0, a particle of water travels along a horizontal elliptical. When we consider surface waves where h is much greater than the wavelength (~1/k), the following approximation can be made: where k is the wave number, h is the distance from mean water level to the bottom and ω is the wave angular frequency. When we consider x0 = 0, equation (1) becomes x = Ax ( z 0 ) cos ωt , z = z 0 − Az ( z 0 ) sin ωt Ax = Az ≈ ae kz 0 (2) Then, a particle of water travels in a circular trajectory with a radius indicated by equation (3). Fig. 2 shows the trajectories of water particles in a surface wave. where ION GNSS 17th International Technical Meeting of the Satellite Division, 21-24 Sept. 2004, Long Beach, CA (3) 1236 Table 1 Expected accuracy of GPS navigation accuracy (3σ) algorithm note horizontal vertical point-positioning 10 m 20 m differential GPS 1m 2m kinematic GPS 5 cm 10 cm < 20km* * Baseline length Table 2 Expected characteristics of GPS system errors error sources range (1σ) time constant ephemeris ~3m ~ 1 hr satellite clock ~3m ~ 5 min ionosphere ~9m ~ 10 min troposphere ~2m ~ 10 min multipath ~3m ~ 100 sec receiver noise ~1m white noise GPS point positioning navigation 120 118 116 x 10 4 power spectrum 114 GPS altitude (m) power spectrum of GPS positioning error band of ocean waves 9 112 110 108 3 2 1 0 -4 10 106 10 -3 10 104 -1 10 frequency (Hz) 10 0 10 1 Fig. 4 Power spectrum of GPS position error 102 0 500 1000 1500 2000 2500 time (sec) 3000 3500 4000 GPS altitude filtered by HPF (m) 100 -2 Fig. 3 Altitude given by GPS point-positioning navigation (experiment) If the movement of a floating buoy corresponds to that of a water particle in an ocean wave, a GPS receiver fixed to the buoy would measure rotational motion with a period of 0.1-20 sec (0.05-10 Hz) as shown in Fig. 2. 1 0 -1 -2 0 500 1000 1500 2000 time (sec) 2500 3000 3500 Fig.5 GPS altitude filtered by high-pass filter with 0.02 Hz cut-off frequency CHARACTERISTICS OF GPS MEASUREMENT moves slowly on account of GPS system errors. This fluctuation indicates the error of GPS navigation by the point-positioning algorithm. However, it should be noted that the time constant of the fluctuation is on the order of from one hundred seconds to several ten minutes order, that is, less than 0.01 Hz. This is because the GPS system errors that affect the positioning error have relatively long time constants. The accuracy of GPS positioning depends on the navigation algorithm, such as the point-positioning algorithm, the differential GPS algorithm and the kinematic GPS algorithm. Table 1 summarizes the expected accuracies of these navigation algorithms. According to the table, only the kinematic GPS method satisfies the centimeter-level accuracy required to sense buoy movement because it eliminates almost all GPS error sources and uses a carrier phase with low receiver noise. On the other hand, the point-positioning algorithm uses pseudorange with larger receiver noise and suffers from the effects of GPS system errors such as atmospheric errors, ephemeris errors, etc. Fig.3 shows a typical point-positioning result over one hour of measurements from a regular GPS receiver set on the roof of a building. Although the GPS antenna is fixed, the positioning result ION GNSS 17th International Technical Meeting of the Satellite Division, 21-24 Sept. 2004, Long Beach, CA 2 Table 2 summarizes the GPS system errors and their expected time constants. As shown in the table, with the exception of receiver noise, the principal GPS system errors have time constants greater than 100 sec. Therefore, the fluctuation of GPS altitude error shown in Fig. 3 moves gradually. Fig. 4 shows the power spectrum of the positioning data of Fig. 3, which is the power spectrum of the error of GPS navigation by the point-positioning 1237 algorithm. As predicted by Table 2, almost all of the power of the GPS positioning error exists in a band less than 0.01 Hz. On the other hand, buoy movement excited by ocean waves is rotational with a period of 0.1-20 sec (0.05-10 Hz), which is also shown in Fig. 4. Therefore, a suitably designed high-pass filter can extract the movement of a GPS equipped buoy excited by ocean waves with minimum influence from GPS point-positioning errors. This is the key principle of measuring ocean waves by GPS point-positioning. H ( z) = H ( e jω T ) = θ (ω ) = −τ c ω θ (ω ) ω = −τ c τ p (ω ) = − τ g (ω ) = − (4) (10) This means the delay of the envelope of waves coincides with that of each wave. This is an important property that our high-pass filter should have to reconstruct the waveforms correctly. Cut-off frequency fc Fig. 6 shows the amplitude characteristic of a high-pass filter. The cut-off frequency is the width of the window open to ocean waves without any reduction of amplitude. As described in the previous section, the longest period of ocean waves is 20 sec. In this research, we keep some margin and set the cut-off frequency to 0.03 Hz. (5) ∞ ∑ x(kT )h(nT − kT ) k = −∞ where h(nT) is called the impulse response. Through the z-transform, equation (5) becomes ION GNSS 17th International Technical Meeting of the Satellite Division, 21-24 Sept. 2004, Long Beach, CA dθ (ω ) dω (9) = −τ c k = −∞ Y ( z) = X ( z)H ( z) (8) In this case, the phase delay and group delay become the same form. ∑ = (7) Linear phase characteristic When the phase characteristic is expressed as below, the time invariant system has a linear phase characteristic. ∞ y (nT ) = ℜ x(kT )δ (nT − kT ) k = −∞ ∑ x(kT )ℜ[δ (nT − kT )] − jωnT When designing a high-pass filter for measuring ocean waves, we considered the following. where δ is the unit impulse function. We construct the high-pass filter by the linear time invariant system. Then, the output of the filter is expressed as follows: ∞ ∑ h(nT )e We can then design the impulse response, which determines the parameters of the digital high-pass filter, by using filter characteristics such as amplitude characteristic, phase characteristic, phase delay and group delay. k = −∞ = ∞ n = −∞ By limiting the bandwidth of the GPS point-positioning data to cut out positioning error, a high pass filter can extract the form of ocean waves correctly with minimum effect of GPS system errors. The filter’s input data are the sequence of sampled positions of the GPS antenna on a buoy, and are generally described as follows: ∞ −n When we replace the variable z by e − jωT , we obtain the frequency response of the transfer function H: DESIGN OF HIGH PASS FILTER 3) ∑ x(kT )δ (nT − kT ) ∑ h(nT ) z n = −∞ To demonstrate this, we applied a simple high-pass filter (2nd-order Butterworth) with a cut-off frequency of 0.02 Hz to the data of Fig. 3, and obtained the result shown in Fig. 5. This reduced the fluctuation to 8 cm (1σ). When the high-pass filter is adopted, the mean value of the antenna’s altitude becomes zero; however this is not a problem since we do not have to measure the altitude of the buoy, only track its rotational movement. This result suggested that our idea would work well for measuring ocean waves with centimeter accuracy. x(nT ) = ∞ Edge frequency of blocking region fa This edge frequency of blocking region is the maximum frequency at which the effects of GPS system errors are blocked completely. GPS system errors exist in the region below 0.01 Hz as shown in Fig. 4. Therefore, we set the (6) 1238 1.0 50 Ac Gain (dB) Gain 0 -50 -100 -150 Aa 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 Frequency (Hz) 0.7 0.8 0.9 1 Phase (deg) 10000 0 0 fa fc 0 -10000 -20000 -30000 Frequency Fig.6 Amplitude characteristic of linear phase high-pass filter Fig. 7a Frequency characteristics of HPF in linear scale 50 Gain (dB) edge frequency to 0.01 Hz. Ripple of pass region Ac As described below, we can not design a filter with ideal characteristics because of limitations of the order of filter parameters. Consequently, there is a ripple in the filter pass band. We set the ripple to less than 0.08 dB (= 1.009), which means that we will have a 9 cm height sensing error for a 10 m high wave. 0 -50 -100 -150 Phase (deg) 5000 0 -5000 -10000 -15000 -20000 10 -4 Suppressed gain Aa There is a positioning error of several meters due to GPS system errors in the region below 0.01 Hz as shown in Table 2. We suppress this error to the order of centimeters by setting the suppressed gain to 40 dB. 10 -3 10 -2 Frequency (Hz) 10 -1 10 0 Fig. 7b Frequency characteristics of HPF in log scale motor DC gain The DC gain is set to zero because absolute altitude data from GPS are useless for measuring wave data correctly. 1.7 m GPS antenna (pendulous) rotational arm counter weight GPS receiver We realize the above performance by designing a linear phase FIR (Finite Impulse Response) filter whose number of the impulse response h(nT) in equation (6) is limited by N ( 0 ≤ n ≤ N − 1 ). There are several methods to determine the appropriate impulse response according to the design conditions, such as the window function method and the Remez method. We adopted the linear programming method to realize arbitrary amplitude characteristics. Fig. 7 shows the design result of our high-pass filter. The order N of the filter is 241. This filter satisfies all of the design conditions. Fig. 8 Schematic view and photograph of test apparatus ION GNSS 17th International Technical Meeting of the Satellite Division, 21-24 Sept. 2004, Long Beach, CA 1239 GPS raw East (m) non jump East (m) To prove the concept, we carried out a laboratory test in which our high-pass filter was applied to observed GPS data. Fig. 8 shows the test apparatus that consists of a wave simulator and a GPS receiver. The wave simulator has a rotating arm to which a GPS antenna is fixed. The rotating arm simulates the motion of a buoy floating in the ocean as shown in Fig. 2. The diameter of rotation is set by an arm length (170 cm), which simulates wave height, and rotation speed can be controlled by varying the speed of the motor. A commercial GPS receiver produced by Furuno Electric Co. Ltd was used. This has 11 channels to track the C/A code and the carrier phase, and carries out point-positioning navigation using all GPS satellites in view. The output rate of navigation data is 5 Hz. 4 2 0 -2 -4 4 2 0 -2 -4 filtered East (m) PROOF OF CONCEPT BY LABORATORY TEST 2 1 0 -1 -2 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 0 200 400 600 time (sec) 800 1000 1200 GPS raw North (m) non jump North (m) In the experiment, we collected GPS point-positioning data for 20 min, which is widely used as a sensing period for ocean waves. We set the period of rotation to 11 sec and the direction of the arm to 114 deg from the true north. Fig. 9 depicts the GPS data in vertical and horizontal planes. The origin of the graph is the initial GPS position. As shown in the figure, the point-positioning data drifted gradually on account of GPS system errors although the wave simulator did not change its position. It should be also noted that there is a jump in the GPS data around 6,550 sec UTC. This kind of position jump is often seen in GPS receivers, and is due to satellite change, carrier phase lock-loss, etc. 4 2 0 -2 -4 4 2 0 -2 -4 filtered North (m) Fig. 10a Result of processed GPS data in east axis 2 1 0 -1 -2 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 0 200 400 600 time (sec) 800 1000 1200 GPS raw Height (m) 4 2 0 -2 -4 non jumped Height (m) 4 2 0 -2 -4 filtered Height (m) Fig. 10b Result of processed GPS data in north axis 2 1 0 -1 -2 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 0 200 400 800 1000 1200 600 time (sec) Fig. 10c Result of processed GPS data in height axis Table 3 Statistics of wave data derived from GPS parameter experiment result true value* wave height 170 cm 167.5 cm wave period 11 sec 10.83 sec wave direction 114 deg 115.7 deg * True values are measured by other means such as measure, clock and magnetic compass. ION GNSS 17th International Technical Meeting of the Satellite Division, 21-24 Sept. 2004, Long Beach, CA 1240 Num. of GPS mean sea level 7 6 5 5600 2 5800 6000 6200 6400 6600 6800 7000 H1 H2 Height (m) 1 0 -1 T1 T2 -2 -3 5600 5800 6000 6200 6400 UTC time in day (sec) 6600 6800 7000 Fig. 11 Definitions by zero up cross method 1.4 estimated period is 0.17 sec, and the error of the estimated direction is 1.7 deg. These results confirmed our proposed concept for measuring wave height and direction using a low-cost point-positioning GPS receiver. 1.2 1 North (m) 0.8 PARAMETERS OF OCEAN WAVES 0.6 0.4 An actual ocean wave is a superposition of various wind waves and swells. The waveform is not a simple sine curve given by the previous ground test. Therefore, we must define the key ocean wave characteristics - height, period and direction - before analyzing the GPS data. The zero up cross method is widely used to define wave height and period. Fig. 11 shows an example of an observed waveform. In this method, the waveform is delimited at the up crossing point of a water surface, and each interval is considered as a single wave. The height of a single wave is the distance between the top and the bottom of the wave. The period is defined to be a length of the interval. 0.2 0 -0.2 -1.5 -1 -0.5 East (m) 0 0.5 1 Fig. 9 Position data of GPS installed on wave simulator We performed the following process to extract the buoy motion (in this case, the rotational motion of the GPS antenna on the wave simulator). The first step is to change the coordinate system from latitude, longitude and height to an east, north and height local frame. The origin of the local frame is the initial position from GPS because absolute position is not necessary to obtain the rotational motion of the buoy. The second step is to remove jumps in the GPS position data. As previously described, jumps are often seen, and their higher harmonics become a problem when using a high-pass filter. We therefore remove jumps as a second step using a simple algorithm, where we remove the offset between the positions before and after each jump. To detect jumps, we check for extremely large changes in position as well as satellite changes for navigation. After the second step, we apply the high-pass filter to the pre-processed GPS data. Fig. 10 shows the position results of the above three steps in east, north and height axes respectively. You can see the lower graph, which shows the output of the third step indicates the antenna’s rotational motion precisely. Table 3 summarizes the statistics of the processed data. The mean wave height (diameter of rotation) is 1.675 m, a deviation of only 2.5 cm from the true value. The error of the ION GNSS 17th International Technical Meeting of the Satellite Division, 21-24 Sept. 2004, Long Beach, CA In the analysis, we delimit the waveforms and make a group of single waves. The number of waves is defined as the result of counting single waves in the group during the observation period. The wave height and period are usually defined in the concept of significant wave, 1/10 maximum wave and maximum wave. The wave height and period of the significant wave are the mean of those of waves extracted by the criterion of wave heights from maximum to one third ranked in descending order. The 1/10 maximum wave depends on the criterion from maximum to one tenth ranked in descending order. The maximum wave is a single wave that has the maximum height. There are many definitions for wave direction but most of these depend on the spectral analysis of ocean waves. We define the movements of a buoy in the east, north and height axes as xe, xn and xh respectively. The correlation functions of these are defined as follows: 1241 1 T →∞ T Aij (τ ) = lim ∫ T /2 −T / 2 1 a 0 + (a1 cos θ + b1 sin θ ) 2 + (a 2 cos 2θ + b2 sin 2θ ) + L xi (t ) x j (t + τ )dt , i, j = e, n, h (11) G (ω , θ ) ≈ The cross spectra are derived from the Fourier transform of equation (11) C ij (ω ) = K ij (ω ) − jQij (ω ) = ∫ ∞ −∞ Aij (τ )e jωτ dτ The Fourier parameters in equation (16) and the cross spectra in equation (14) have the following relationships: (12) a1 = Because the movement of the buoy tracks a circle as shown in Fig. 2, the phase difference between xh and xe, xh and xn is 90 deg, and the phase difference between xe and xn is 0 deg or 180 deg. We then have the following equation: K he = K hn = Qen = 0 b1 = b2 = (13) (17) 2 K en φe + φ n We performed an ocean field test using a point-positioning GPS receiver installed on a buoy in the Kitan Strait between the islands of Honshu and Shikoku from June 17 to June 23, 2004. Fig. 12 shows a map indicating the Kitan Strait and a photograph of the strait and the buoy. In addition to the GPS receiver, the buoy was equipped with an accelerometer based wave sensor, and an ultrasonic wave sensor was located on the seabed in the vicinity but about 5 km away from the buoy. Data from these sensors were used as reference data to evaluate the GPS-based wave sensing system. C hn = − jQhn (14) C nn = φ n C hh = φ h where φe, φn, φh denote the power spectrum of the wave in each axis, and φh in particular is the power spectrum of the wave φ. The spectral resolution of ocean wave energy can be expressed by multiplication of the power spectrum and its direction distribution function: S (ω , θ ) = φ (ω )G (ω , θ ) φh Qhn FIELD EXPERIMENT RESUTLS C he = − jQhe C ee = φ e Qhe φh φ − φn a2 = e φe + φ n After substitution of equation (13), equation (12) becomes as follows: C en = K en (16) Then, we can calculate the direction distribution function (15) from the cross spectra. The wave direction θ is defined where θ is the direction of the wave. Each function has following relationships: as the direction in which G(ω, θ) has the greatest value ( ∂G / ∂θ = 0 ). 2π φ (ω ) = ∫ S (ω , θ )dθ θ = tan −1 0 2π ∫ G(ω ,θ )dθ =1 0 (18) We usually calculate the wave direction at the frequency ω that has the first, second and third greatest power spectra. According to Longuett-Higgins, Cartwright and Smith (1963)5), the direction distribution function can be expanded as a Fourier series: ION GNSS 17th International Technical Meeting of the Satellite Division, 21-24 Sept. 2004, Long Beach, CA Q b1 = tan −1 hn a1 Qhe 1242 14 accelerometer ultrarsonic GPS 12 wave period (sec) 10 Kitan Strait 8 6 4 Buoy 2 Ultrasonic sensor 0 00:00 12:00 00:00 12:00 00:00 12:00 00:00 12:00 00:00 12:00 00:00 12:00 00:00 12:00 00:00 06/17 Fig.13b 06/18 06/19 06/20 06/21 date/time (UTC) 06/22 06/23 Significant wave period observed at the Kitan Strait 350 Fig. 12 Kitan Strait and buoy equipped with GPS receiver accelerometer ultrarsonic GPS 300 250 7 200 direction (deg) accelerometer ultrarsonic GPS 6 5 150 wave height (m) 100 4 50 3 0 00:00 12:00 00:00 12:00 00:00 12:00 00:00 12:00 00:00 12:00 00:00 12:00 00:00 12:00 00:00 2 06/17 06/18 06/19 06/20 06/21 date/time (UTC) 06/22 06/23 Fig. 13c Wave direction at frequency that has maximum power spectrum, observed at the Kitan Strait 1 0 00:00 12:00 00:00 12:00 00:00 12:00 00:00 12:00 00:00 12:00 00:00 12:00 00:00 12:00 00:00 06/17 06/18 06/19 06/20 06/21 date/time (UTC) 06/22 06/23 In the analysis, we used 20 min GPS point-positioning data each hour during the experiment. Fig. 13 shows the results of the analysis, which show the height and period of significant waves and wave direction at the frequency with the biggest power spectrum. During the experiment, the sixth typhoon of the season passed close to the Kitan Strait as shown in Fig. 14. Our system therefore experienced ocean waves from small wind waves to long swells. When the typhoon passed, our system observed that the significant wave height reached 7m where the maximum wave height was over 10 m. Fig. 15 shows the reconstructed trajectory of the buoy movements in the east, north and height axes when the maximum wave height was observed. Fig. 13a Significant wave height observed at the Kitan Strait ION GNSS 17th International Technical Meeting of the Satellite Division, 21-24 Sept. 2004, Long Beach, CA 1243 Because the GPS wave sensor and accelerometer wave sensor were installed on the same buoy, the GPS derived data coincide very well with those of the accelerometer. The differences between the two systems are summarized in Table 4. Table 4a shows the statistical result for all data during the experiment and Table 4b indicates the results corresponding to significant wave heights. Table 4b is indicated because the error of the accelerometer based wave sensor is proportional to wave height. The accelerometer sensor has a guaranteed 10 % accuracy in this case, which means a 70 cm error is expected for a 7 m significant wave height. The differences shown in Table 4b coincide well with the accelerometer error specifications, which confirms the correctness of the GPS wave sensor. On the other hand, the data from the ultrasonic wave sensor shows smaller values. This is because the ultrasonic sensor was located apart from the buoy in the neighborhood of the coast that may have reduced the energy of the waves. Wave direction obtained from GPS corresponds well to the ultrasonic sensor in Fig. 13, where the resolution of the wave direction data is 10 deg and 22.5 deg for the GPS and ultrasonic sensor respectively. However, the accelerometer wave sensor data differs from these. It may be difficult to keep the 3-axis accelerometers horizontal, and in that case that the data of the accelerometer in horizontal plane would be unreliable. This would cause the drift seen in the accelerometer wave direction data. From this ocean experiment, we can conclude that the concept of GPS wave sensor is verified and that the GPS wave sensor can replace existing accelerometer based and ultrasonic based wave sensors. buoy movement H (m) buoy movement N (m) buoy movement E (m) Fig. 14 Trajectory of sixth typhoon (2004) during experiment 4 0 -4 -8 8 4 0 -4 -8 8 maximum wave height 10.1 m 4 0 -4 -8 0 1:4 2:4 0 2:0 2:4 0 2:2 2:4 0 2:4 2:4 0 0 0 0 0 0 0 3:0 3:2 3:4 4:0 4:2 4:4 5:0 2:4 2:4 2:4 2:4 2:4 2:4 2:4 UTC time (June 21, 2004) Fig. 15 Reconstructed waveform by GPS Table 4a Differences between the GPS and accelerometer wave sensors GPS BASED WAVE SENSOR BOARD data wave height wave period The new GPS based system does not require complicated processing to extract wave data, but simply applies a high-pass filter to point-positioning data. Therefore, the algorithm can be executed in the navigation processing core of any GPS receiver. We developed a GPS receiver board shown in Fig. 16, which includes wave height and direction measurement functions. This GPS board enables us to reduce the amount of data transmitted from sensors to a land station via communication satellite because it is not necessary to transmit raw GPS data for processing, only analyzed ocean wave data is transmitted. This means that our method reduces not only the price of a sensor system but also the operating cost. ION GNSS 17th International Technical Meeting of the Satellite Division, 21-24 Sept. 2004, Long Beach, CA 8 mean 8.8 cm 0.6 sec 1σ 27.6 cm 0.8 sec rms 28.9 cm 1.0 sec Table 4b Wave height differences between the GPS and accelerometer wave sensors corresponding to significant wave height height range mean 1σ rms 0-1 m -4.3 cm 6.8 cm 8.1 cm 1-3 m 11.4 cm 20.0 cm 22.9 cm 3-5 m 43.6 cm 16.7 cm 46.8 cm >5 m 50.8 cm 76.7 cm 92.1 cm Note: The accelerometer sensor has a guaranteed 10 % accuracy in wave height. 1244 CONCLUSIONS We propose a new method to measure wave height and direction by installing the point-positioning GPS receiver on buoys placed in the open sea. The essential principle of measuring wave height with centimeter accuracy depends on the characteristic of GPS measurement errors which allows them to be removed by a sophisticated high-pass filter without affecting wave information. Laboratory tests confirm the validity of proposed method and the GPS based wave sensor shows an accuracy of several centimeters in wave height and less than 1 sec in wave period. Wave direction can be also measured with an accuracy of several degrees. Experiments at sea verify the feasibility of our proposed system, and the experiment data indicate the GPS based wave sensor can replace existing wave sensors. Fig.16 GPS board including wave sensing function ACKNOWLEDGMENT We would like to appreciate the great assistance of Kinki Trunk Roads Investigation Office, Kinki Regional Development Bureau of the Ministry of Land, Infrastructure and Transport for field experiments at the Kitan Strait. REFERENCES 1) 2) 3) 4) 5) Tatsuo Tokuoka, Theory of Wave Motion, Science Co. Ltd., Tokyo, 1984 (in Japanese). J. Lighthill, Waves in Fluids, Cambridge University Press, Cambridge, 1978. Hiroshi Ochi, Digital Signal Processing Studied by Simulation, CQ Press Co. Ltd., Tokyo, 2001 (in Japanese). J. H. McClellan, T. W. Parks and L. R. Rabiner, “A Computer Program for Designing Optimum FIR Linear Phase Filters”, IEEE Trans, Audio Electro., Vol. 21, pp. 506-526, 1973. M. Longuett-Higgins, D. E. Cartwright and N. D. Smith, “Observations of the Directional Spectrum of Sea Waves Using the Motions of a Floating Buoy”, Ocean Wave Spectra, U. S. Naval Oceanographic Office, 1963. ION GNSS 17th International Technical Meeting of the Satellite Division, 21-24 Sept. 2004, Long Beach, CA 1245