Pilot Study on the use of Quasi-Zenith Satellite System as a GNSS Augmentation System for High Precision Positioning in Australia

International Global Navigation Satellite Systems Society IGNSS Symposium 2015 Outrigger Gold Coast, Qld Australia 14-16 July, 2015 Pilot Study on the use of Quasi-Zenith Satellite System as a GNSS Augmentation System for High Precision Positioning in Australia Ken Harima School of Mathematical and Geospatial Sciences, RMIT University, Australia Phone: +61 3 9925 3775 Fax: +61 3 9663 2517 Email: ken.harima@rmit.edu.au Suelynn Choy School of Mathematical and Geospatial Sciences, RMIT University, Australia Phone: +61 3 9925 2650 Fax: +61 3 9663 2517 Email: suelynn.choy@rmit.edu.au Mohammad Choudhury School of Civil and Environmental Engineering, University of New South Wales, Australia Phone: +61 2 9385 4173 Fax: +61 2 9385 5657 Email: mohammad.choudhury@unsw.edu.au Satoshi Kogure Satellite Application and Promotion Centre, Space Applications Mission Directorate, Japan Aerospace Exploration Agency, Japan Phone: +81-50-3362-2456 Fax: +81-29-868-5987 Email: kogure.satoshi@jaxa.jp Chris Rizos School of Civil and Environmental Engineering, University of New South Wales, Australia Phone: +61 2 9385 4205 Fax: +61 2 9385 6139 Email: c.rizos@unsw.edu.au ABSTRACT The Quasi-Zenith Satellite System (QZSS) is a Japanese regional satellite navigation system covering East Asia and Oceania. QZSS is designed as a satellite-based augmentation system to improve the availability and performance of Global Navigation Satellite Systems (GNSS) in the region. QZSS transmits augmentation signals such as the LEX (L-band Experimental, to be renamed L6 for operational QZSS) signal, which is capable of transmitting precise navigation messages for high accuracy point positioning. A joint project between the Australian Cooperative Research Centre for Spatial Information (CRCSI) and the Japanese Aerospace Exploration Agency (JAXA) aims to assess the feasibility of using the LEX signal to potentially deliver a precise positioning service in Australia. This paper presents the results of a pilot study for the use of the LEX signal to transmit navigation messages for Precise Point Positioning with Ambiguity Resolution (PPP-AR). LEX correction messages based on the available realtime products were transmitted from Australia to the QZS master control station in Japan and broadcasted on the LEX signal. The messages were decoded on different points regions in Australia and used to compute realtime PPP-AR solutions. Positions solutions were obtained for both fixed receivers and vehicle mounted receivers. A comparison with Position solutions obtained using real-time streams directly shows that latency and outages product of the satellite transmission don t have significant effects in ambiguity resolution nor positioning accuracy. KEYWORDS: Real-time PPP, Ambiguity resolution, QZSS, GNSS Augmentation

1. INTRODUCTION The Quasi-Zenith Satellite system (QZSS) is a Regional Navigation Satellite System (RNSS) in development by Japan. The development of QZSS is planned to have three phases. In the current first phase the QZS-1 satellite, launched on the 11th of September 2010 (Kishimoto et. al., 2011), is used to validate the technology, study potential applications and identify possible utilization challenges. During the second phase scheduled to start on 2018, the number of satellite will be increased to four (Kallender-Umezu,2013). At this stage the QZSS will serve as a GNSS augmentation system that covers the East Asia and Oceania Region. The full constellation consisting in 7 satellites is expected to be available by the end of 2023. The QZSS transmit a number of performance enhancement signals in order to fulfil its role as a GNSS augmentation system. One such signal is the LEX signal, which is designed to transmit correction messages that enable position, navigation and timing applications requiring centimetre-level accuracy. The LEX signal is transmitted using the same carrier frequency as the Galileo E6 signals (1278.75 MHz). The Centimetre-Level Accuracy System (CLAS) has been tested in Japan and demonstrated its capability of providing accuracies of a few centimetres with convergence times of less than one minute (Saito et. al, 2011). However since these messages depend on local tropospheric and ionospheric measurements that rely on Japan s network of Continuously Operating Reference Stations (CORS), the CLAS system will only be available in or near Japan. Complementary systems that are valid over the other parts of the region are also in development by the Japanese Aerospace Exploration Agency (JAXA). The development of these systems is based on Precise Point Positioning (PPP). JAXA s system for the LEX signal have been tested in Japan achieving decimetre level positioning within a few hours (Suzuki et. al, 2014). A joint research between the Australian Cooperative Research Centre for Spatial Information (CRCSI) and JAXA was established to test the feasibility of utilising the QZSS LEX signal to deliver high accuracy real-time precise positioning in Australia. The transmission of regional or national messages for precise positioning using the QZSS LEX signal is of interest for Australia as it will enable accurate positioning on parts of the country in which communication networks are either unavailable or unreliable. The present paper gives a brief description of the activities undertaken to assess the potential of the QZSS LEX signal to deliver high precision positioning in Australia. The research project focuses on Precise Point Positioning (PPP) as a positioning technique. PPP as a technique focuses on obtaining precise solutions using a single GNSS receiver without the need to communicate directly to a nearby reference station. For the afore mentioned reasons, GNSS augmenting information for PPP, delivered using the QZSS LEX signal has most realistic potential in delivering high accuracy positioning to remote regions in Australia. The rest of the paper is structured in four sections. A brief description of the Japanese QZSS system and the LEX signal is given in section 2. The PPP technique and its potential performance is described in section 3. Preliminary tests of PPP using the QZSS LEX signal, including real-time tests using JAXA s PPP messages and the transmission of messages from Australia are presented in section 4. Finally, a summary of findings and plans for future research are outlined in Section 5.

2. QZSS and the L6 signal The following section describes the characteristics of the QZSS and its L6 signal, in particular as part of its functionality as a GNSS augmentation system. The QZSS will function as a GNSS augmentation system for the East-Asia and Oceania region by 2018. The LEX signal studied in this research is the experimental version of a signal to be named L6 once the QZSS is operational. The L6 signal, having higher data rates than usual SBAS signals, has great potential as a channel to deliver corrections for high accuracy GNSS positioning. 2.1 QZSS satellites and signals The QZSS is by design a RNSS that covers the East-Asia and Oceania region. In order to obtain this area of coverage, the QZSS will use a combination of Geostationary (GEO) and (Geosynchronous) Highly-inclined Elliptical Orbits. The QZSS elliptical orbits are designed in a way that satellites placed in such an orbit will stay near the zenith on Japan for at least 8 hours a day, for this reason they are also called Quasi-Zenit Orbits (QZO). The red line in Figure 1 shows the ground track for the QZO satellites, while the black marker shows the nominal position of the first GEO satellite. Currently the QZSS consist of one satellite, the QZS-1 satellite placed in this QZO orbit. The nominal orbit parameters for the four satellites to be active starting in 2018 are show in Table 1. Parameter QZO-1 QZO-2 QZO-3 GEO-1 Semi-major axis 42164 Km 42164 Km 42164 Km 35786 Km Eccentricity 0.075 0.075 0.075 0 Inclination 40 40 40 0 Argument of Perigee 270 270 270 N/A Longitude of Ascending Node 117 247 347 N/A Central Longitude 136 E 136 E 136 E 127 E Table 1. Orbital Parameters of QZSS satellites to be launched by 2018 The orbits of QZS-2 and QZS-3 are scheduled to be put in orbits that pass over the same place on the earth s surface as the QZS-1 but with a time delay of 8 hours and 40 minutes (the Longitude of Ascending Node and Mean Anomaly are displaced 130 with respect to QZS-1). It is expected that by 2023 a QZS-4 satellite will replace QZS-1 and QZS-5 will complete a set of four satellites with similar ground track with time spacing of around 6 hours. As a GNSS augmentation system, the QZSS aims to provide availability and performance enhancement to GNSS systems in the region of coverage. Availability enhancement will be achieved by transmitting signals that are compatible with GPS satellites, i.e. L1 C/A, L2C, L5 and L1C signals, effectively increasing the number of GPS satellites available in the region. Performance enhancement will be achieved by transmitting two additional signals, i.e. L1S and the L6 signal (JAXA, 2014), which has the objective of enabling high accuracy positioning by transmitting correction and monitoring information on the GNSS signals available in the area. The L1S signal is currently named L1-SAIF for Sub-meter class Accuracy with Integrity Function, and as this name implies it is meant to deliver decimetre level precision with integrity assurance features. The augmentation information transmitted have a format that is similar to well established SBAS satellites like WAAS and EGNOS. When the QZSS is functional as an Augmentation system, the L1S signal will be used to complement and eventually replace the MSAS system.

The L6 signal is unique to the QZSS system and is proposed as a message broadcasting channel for positioning augmentation for applications requiring centimetre-level accuracy. Although similar messaging channels like the E6 signal from Galileo and the B3 signal from Beidou systems are been deployed, the QZSS uniqueness comes from its Code Shift Keying modulation, which enables higher data rates than similar channels using binary shift keying modulation. Table 2 show the carrier frequency and data rates for the different messaging channels available for GNSS augmentation. The characteristics of the Beidou B3 signal is not yet available to the public. Message Constellation Carrier (MHz) Data Rate (bps) NAV GPS 1575.42 50 SBAS WAAS/MSAS/EGNOS 1575.42 250 C/NAV Galileo 1278.75 500 L6 (LEX) QZSS 1278.75 2000 Table 2. Characteristics of different GNSS augmentation signals 2.2 L6 Signal coverage Although at 2000bps, the QZSS L6 signal will boast about 8 times the data rate of traditional SBAS signals, there is a price to pay. Although the average signal power for the L6 signal is set to be about 2.5 times higher than the L1S signal, this still means that there is less than a third of the energy available per data bit. For this reason the L6 signal will impose higher requirements in the receiver antenna gain. Figure 1. Regions in East-Asia with QZSS L6 availability. The L6 signal imposes stricter restrictions on receiver antennas. With a survey grade antenna the L6 signal can be received as long as the satellite is above 20 of elevation (right), with a standard patch antenna it needs to be at 40 (left). In a previous publication, the authors presented the success rate of a software receiver (Lighthouse Ltd. s Lex Message Streamer) using an standard patch antenna. It was shown that the QZS-1 satellite had to be over 40 of elevation for the signal to be demodulated and decoded reliably. Similar tests using a Trimble choke ring as an antenna showed that using

this type of antenna the elevation requirements reduces to 20. Figure 1 shows the potential area of coverage for the QZSS L6 signal assuming 20 (left) and 40 elevation is required (right). 3. PPP for stations in Australia While the QZSS area of coverage and the data capacity of the L6 signal gives a great potential for high accuracy positioning in Australia and New Zealand, the positioning techniques that the messages support must also be adapted to the task of delivering positioning services over a vast area with concentrated mobile communication and GNSS monitoring infrastructure. The following section will describe the PPP algorithm and describe the potential performance of the method in Australia. 3.1 Precise Point Positioning Traditional PPP uses an iono-free linear combination of measurements from two different carrier frequencies to eliminate the effects of the ionospheric delay (Zumberge, 1997), which is the largest error with local variability: (1) here f i is the frequency of the L i band carrier and P i and φ i are the satellite single-differenced pseudorange and carrier phase measurement for the signal on the L i band. (2) where ρ is the estimated satellite-receiver distance, dt is the satellite clock error, T is the tropospheric delay, c is the speed of light, N i is the phase ambiguity, and B Pi and B Li correspond to the satellite biases for the pseudorange and carrier phase measurements. Index for satellite is omitted for simplicity. Since the ionospheric delay can be considered to be inversely proportional to the square of the carrier frequency, they are eliminated by forming the linear combination in (1) leaving only tropospheric delays, satellite biases, satellite orbit and satellite clock as sources of errors. (3) The tropospheric errors can be estimated with centimetre level accuracy based on the tropospheric zenith delay, and the other tree sources of error are global in nature, thus they can be measured using a sparse network of global CORS. In its most basic form, the messages transmitted to the user are precise orbits and iono-free satellite clock corrections (4) By correcting the pseudorange and carrier phase measurements using the external satellite clock correction equation (3) becomes

where (5). (6) The end user then will estimate the sum of the last three components of equation (5) as a floating ambiguity along with the rover position and the zenith tropospheric delay. The horizontal positioning accuracies for post-processed PPP solutions are illustrated in figure 2. The positioning solutions presented here are the average of positioning solutions calculated based on daily measurements taken from 10 reference stations in Australia: Alice Springs (ALIC), Ceduna (CEDU), Darwin (DARW), Perth (HIL1), Hobart (HOB2), Karratha (KARR), Melbourne (MOBS), Tibooburra (TBOB), Toowoomba (TOOW) and Townsville (TOW2). Figure 2. Horizontal positioning accuracies for standard PPP and PPP-AR (Post processing) The observables from the stations as well as messages from the IGS real-time stream CLK9B were recorded from 12 th to 18 th of August 2014 and 12 th to 18 th September 2014. The blue line in Figure 2 represent the represent the basic float PPP solution. As it can be seen in the graph, the solution took an average of three hour to converge to its steady state accuracy. The 2σ RMS horizontal errors discarding the first three hour of each daily solution was about 6.5 cm. These results, as well as all other presented in this paper are GPS only solutions and were generated using a modified version of the RTKLIB software (Takasu, 2009). It is to note that the convergence time of individual results show large variations as a result of the differences in the number of visible satellites and their geometry. 3.2 PPP with ambiguity resolution The next step in improving the accuracy of PPP solutions and their convergence times is to allow the resolution of the integer carrier phase ambiguities N i. By solving and fixing the values of N i to integer values, error associated with these parameters can be eliminated. This combined with the transmission of precise values for satellite orbits, clocks and biases, allows the use of carrier phase measurements to calculate the receiver position with higher accuracy.

Several strategies have been proposed for PPP-AR, each associated with specific types of corrections transmitted to the end-user (Geng, 2010. Collins, 2010). As an example the method proposed by Laurichesse (Laurichesse et. al., 2009 & 2011) from French Space Agency CNES, solves the equations in (2) directly estimating T and I as variables. On the other hand, other PPP-AR strategies like those prosed by Ge. (Ge et. al, 2007) limits itself to resolving the ambiguity of a single differenced, ionosphere-free combination like those in equation (3). We adopted this method for demonstration purposes as it becomes a natural extension of the basic PPP algorithm. Additional information is required to be sent to the user in order to allow the resolution of these ambiguities. The iono-free phase bias, B if in (6), will be sent to the user. By correcting the carrier phase measurements with this bias, the carrier phase will become a function of only rover position, Tropospheric corrections and the integer biases. In Ge s Method, the Melbourne-Wubenna combination is used as an extra measurement to recover the integer nature of the ambiguities: were (7) transmitting the Melbourne-Wubenna bias B MW, allows the user to solve the N 1 -N 2 ambiguities and recover the integer nature of ambiguities in (5). An example of ambiguity resolved PPP solutions using measurement from the MOBS station is shown in Figure 3. Like the case in Figure 3, about 80% of solutions, converge to a fixed solutions within one and half hours. There are instances however in which solutions took more than 3 hours to achieve ambiguity resolution, this large range of convergence times is attributable to variations in geometrical dilution of precision DOP factors. Figure 3. Float PPP and Ambiguity resolved PPP solutions for MOBS station on the 13 of August 2014. Dark blue line represent the Float PPP solution, green lines represent ambiguity fixed solutions. Light blue lines represent times where ambiguities could not be solved The green line in Figure 2 represents the average of solutions calculated in stations around

Australia during the 12 th to 18 th of August 2014 and 12 th to 18 th September 2014 time frame. Again the varying times to achieve fixed solutions makes the average convergence time about 3 hours. Once a fixed solution is achieved, the 2σ RMS horizontal errors are 4.5 cm which represent an 30% reduction in horizontal accuracy. 3.3 PPP enhanced by local ionospheric measurements As shown in the previous subsection, the accuracy of ambiguity fixed PPP can reach centimetre-level. The main issue for the practical use of PPP techniques are the long convergence times to achieve the desired results. Numerous methods are under investigation to reduce convergence times, these include, partial ambiguity resolution (Wang, 2013), dynamic models, triple frequency PPP methods (Lauricheese, 2015) among others. However, one of the most realistic method for obtaining instantaneous ambiguity resolution is still the introduction of external Ionospheric measurements (Rovira-Garcia, 2014). If an external Ionospheric measurement is available to the user, the modified geometry free combination can be used to assist in the estimation of the N 1 and N 2 ambiguities If an accurate Ionospheric delay estimation can be made available to the user along with the geometry-free phase bias, the ambiguities can be resolved in a few minutes as shown in Figure 4. As with Figure 3, the dark blue lines represent the Float PPP solution, the light blue lines represent points where the ambiguities could not be fixed, and the green lines represent the fixed solutions. Figure 4. PPP solutions with high accuracy Ionospheric information. Results were obtained using MOBS station observables for the 13 of August 2014. Blue line represent the Float PPP solution, green lines represent ambiguity fixed solutions. In order to obtain solution presented in Figure 4, a precise Ionosphere was calculated by fixing the position to the known parameters. The Ionospheric corrections where then used to recalculate the kinematic PPP solutions. The results for measurements taken for the 10 stations and two week show that with an accurate enough external estimation of the ionospheric delays the ambiguities of the PPP solutions can be fixed in less than 5 minutes.

This solutions are of course theoretical only, the real feasibility of these product will depend on the generation of precise Ionospheric maps over Australia which remains an open problem. 4. Real-time PPP using the QZSS LEX signal This sections will present a set of real-time PPP tests performed using the QZSS LEX as a transmission medium. Solution obtained using the LEX signal show little difference with respect to the same corrections transmitted through NTRIP caster. MADOCA based messages and CNES based messages can both achieve positioning accuracies around 5 to 6 cm which is significantly better than previously tested GPS only float PPP 4.1 PPP and the MADOCA solutions JAXA s messages for PPP using the QZSS LEX signal are the type 12 messages (JAXA, 2014). These messages are generated using the Multi-GNSS Advanced Demonstration tool for Orbit and Clock Analysis (MADOCA) software. The MADOCA software outputs State Space Representation (SSR) messages for satellite orbits and clocks that follow the Radio Technical Commission for Maritime Services (RTCM) protocol (RTCM, 2013). These messages in turn are packaged into QZSS LEX messages as shown in Figure 5. Figure 5. QZSS L6 messages type 12 (MADOCA messages) Real time evaluation of PPP using corrections from MADOCA-LEX messages have been performed as part of the joint research project between the Australian CRCSI and JAXA. Real-time tests were performed at a fixed reference station and a vehicle experiment was performed on the 22 October 2013. The positioning accuracy was 8.8 cm of horizontal RMS and 14.5cm of 3-dimentional RMS (Choy, et. al. 2015). 4.2 Real-time transmission tests in fixed positions As a benchmark for a system using the QZSS LEX signal to enable high accuracy positioning in Australia, an experiment was performed in which LEX messages generated in Australia were sent to the QZSS MCS to be broadcasted by the QZS-1 satellite. In order to obtain the messages for transmission, messages from the IGS CLK91 stream (IGS, 2013) and CLK9B stream from the PPP-WIZARD project, both generated by the French Space Agency CNES, were used as a base to generate correction messages for Ge s algorithm. These included Satellite orbits, iono-free clock corrections, iono-free phase biases and Melbourne-Wubenna biases. The biases in turn were encoded into L6 messages using the MADOCA format described in Figure 5. Table 3 show the contents of the RTCM messages packaged into the QZSS LEX signal.

Type UDI Data Description O R Orbit offset (Radial) 1057 5 sec O A Orbit offset (Along-track) O C Orbit offset (Cross-track) 1062 5 sec Clock Correction 1059 5 sec B MW (ID:24) MW combination bias B IF (ID:25) Ionosphere-free combination bias Table 3. Contents of the QZSS L6 messages for PPP-AR. The bias ID for B WL and B NL were arbitrary assigned and are not part of the standard message 1059. The broadcasted messages were then received in Australia and used to perform PPP with ambiguity resolution using Ge s Method. Real-time tests in a fixed reference station were performed in June and September 2014. A vehicle test was performed in November 2014. The test site used for the fixed position tests was the GNSS reference station on Bundoora (DORA), Victoria approximately 16Km north of Melbourne. The GNSS receiver, i.e. a Javad DELTA-G3T was connected to the Javad GrAnt-G3T antenna. Both the GNSS receiver and the LEX receiver were established at the RMIT Bundoora tracking station. The approximate coordinates of the location are 37 40 51 S of latitude and 145 03 52 E longitude. Figure 6 shows the time series for the PPP solutions calculated on the 9 th of June 2014. Two solutions were calculated each day using the same GNSS observations. The precise ephemeris information for one of the solutions were decoded in real time from the QZSS LEX signal. For the other solution, the RTCM messages where obtained directly from the CLK9b or CLK91 streams and the necessary bias were calculated inside the rover software. Figure 6. Real-time ambiguity resolved PPP solutions for the 9 of June 2014. Blue line represent the Float PPP solutions, green lines represent ambiguity fixed solutions. In Figure 6 the dark coloured solutions correspond to the solutions obtained using the streams directly whereas the light coloured solutions correspond to solutions obtained using the QZSS

LEX signal as the transmission method. The method of transmission has little influence in the results, indicating that transmission through QZSS LEX is an appropriate method of delivering the PPP corrections (in terms of reliability, latency, etc.). Table 4 and table 5 show the positioning performance of the PPP algorithms during the fixed position tests. Time to fixed solutions varies from 80 minutes to above 3 hour. The 2σ RMS errors of the fixed solutions range from 3.5 cm to 6.2 cm for the horizontal component and 7.0 cm to 14.5 cm for the vertical component. The average value for the horizontal RMS accuracy was 4.9cm, although this is about 10% higher than the post-processing result, it can be easily explained by the use of different receiver and the small sample of real-time solutions. Start day Solution time TTFF Horiz. RMS (2σ) Vert. RMS (2σ) 8 th June of 2014 00:30 to 09:15 UTC 01:32:26 5.0 cm 12.8 cm 8 th June of 2014 22:00 to 04:00 UTC 03:11:58 4.6 cm 14.6 cm 9 th June of 2014 23:00 to 04:15 UTC 02:18:02 5.4 cm 9.8 cm 10 th June of 2014 04:50 to 10:00 UTC 02:22:47 6.2 cm 12.2 cm 27 th Sept. of 2014 17:00 to 21:00 UTC 01:46:54 4.6 cm 7.0 cm 28 th Sept. of 2014 18:30 to 21:30 UTC 01:17:53 4.3 cm 10.0 cm 29 th Sept. of 2014 15:30 to 23:30 UTC 01:38:24 5.8 cm 12.3 cm 30 th Sept. of 2014 18:00 to 23:00 UTC 01:59:03 3.5 cm 12.3 cm 1 st Oct. of 2014 17:30 to 23:30 UTC 02:00:53 4.2 cm 10.3 cm Table 4. Positioning performance of real-time ambiguity resolved PPP using the QZSS LEX signal 4.3 Real-time transmission tests in in a moving vehicle The moving vehicle test presented here was carried out on 12 November 2014 from 13:00 UTC to 17:40 UTC in Centennial Park, Sydney, Australia (2:00 to 6:40 am local time). The GNSS and QZSS LEX antennas were mounted on top of a car roof as shown on Figure 7. Figure 7. Receiver antennas on the roof of the car. The GNSS antenna, a Trimble Zephyr 2 in the back of the car, was used to collect GNSS observables. The LEX antenna, small patch antenna at the front, was used to get precise ephemeris

The antennas then were connected to receivers inside the car. The LEX antenna was connected to a software receiver to demodulate and decode the messages. The GNSS antenna (a Trimble Zephyr 2) was connected to 2 receivers, one of the receivers were a Trimble netr9 receiver from which the GNSS observables for the PPP solutions were obtained. The other receiver was a Leica GS25 used in rover mode to obtain network RTK position solutions to serve as a reference to which to compare the PPP solutions. The ground track followed by the moving car during the experiment is shown in Figure 8. The car was left stationary from 13:00 to 16:16 UTC to let the solution achieve convergence before the driving started. Figure 8. Ground track for the vehicle tests performed on November 2014. The tests Were performed in Centennial Park, Sydney, Australia. Multiple PPP solutions were calculated for the tests using the netr9 GNSS measurements. We present three of those solutions in here. First for the first solution, precise ephemeris from the MADOCA software were transmitted via NTRIP caster from Japan. The MADOCA corrections used this time were an upgraded version from the ones used in October 2013, the key differences being that the they included precise ephemeris for GPS, GLONASS and QZSS satellites, this is in contrast with the tests described in section 4.1, which was a GPS only solution. The other two solutions consisted on precise ephemeris created in RMIT based on data from the CLK91 streams. For one of the solutions, the corrections were transmitted using an NTRIP caster. For the third solution, the LEX messages were transmitted to the QZSS Master Control Station and broadcasted using the QZSS LEX signal. Figure 9 show the time series for the three PPP solutions with respect to the RTK solutions. The blue line represent the Multi-GNSS MADOCA solution, the green line represent the solutions using the RMIT products through the LEX signal. The black line represent the solutions obtained using the RMIT products delivered through an NTRIP caster. As it can be seen from Figure 9, the RMIT through NTRIP had to be restarted after a long outage in the mobile network based connection, presumably on the on the NTRIP caster side.

Figure 9. Time series for vehicle tests in Sydney, 12 November 2014. Blue line represent a Multi- GNSS MADOCA solution, Green line represent the LEX signal delivered RMIT solution, and the black line represent the NTRIP delivered RMIT solution. It is of note that in spite of starting with 1hour and 50 minutes of difference the RMIT through NTRIP and through satellite achieved fixed solutions within 6 minutes of each other. We suspect this is due to the fact that the number of visible during the 13:00 to 14:30 period ranged from 5 to 7 with a GDOP ranging from 2.3 to 3.0. This is in contrast with the 14:30 to 16:30 period when the number of visible GPS satellites was 8 to 10 and the GDOP had a maximum of 2.1. The approximate start time, the time to a stable fix solution, the Horizontal accuracy during the driving test (16:16 to 17:50 UTC) and the Vertical accuracy during the driving test is shown in Table 5 Solution Start time TTFF Horiz. RMS (2σ) Vert. RMS (2σ) Comments MADOCA 13:00 UTC - 5.9 cm 11.3 cm GPS+GLO+QZSS RMIT LEX 13:00 UTC 3:15:20 4.9 cm 13.2 cm GPS Only PPPAR RMIT NTRIP 14:50 UTC 1:31:06 6.3 cm 13.8 cm GPS Only PPPAR Table 5. Positioning performance of real-time PPP different sources of precise ephemeris. The long convergence time for the solution using LEX signal is product of high DOP factors arising from a low number of GPS satellites available between 13:00 and 14:30. This can also explain why having started at different times the solution using RMIT corrections from an NTRIP caster achieved convergence within minutes of the solution using the LEX signal. The horizontal accuracy during the driving tests were 4.9 cm RMS for the RMIT solution and 5.9 cm RMS for the MADOCA solution. The measured accuracy of the RMIT solution is consistent with the accuracies measured in fixed point solutions and the post-process result from section 3.2. Both solutions are significantly more accurate than the GPS only float PPP solutions mentioned in section 4.1 show casting the advantages of Multi-GNSS PPP and ambiguity resolution.

5. SUMMARY AND FUTURE WORK The QZSS is a Regional Navigation Satellite System being developed by Japan. By 2018, with 4 satellites in orbit it will function as a GNSS augmentation system covering most of East-Asia and Oceania, including Australia and New Zealand. The QZSS augmentation signals in particular the L6 (currently known as LEX) signal have a great potential for delivering high accuracy positioning in Australia. GNSS correction information for PPP transmitted using this medium will be able to deliver centimetre-level accuracy positioning for the whole coverage area. In the research presented in this paper, the potential of QZSS L6 delivered PPP was assessed through a series of realtime tests in which Australian generated and Japanese generated corrections for PPP were transmitted using the QZSS LEX signal and used in Australia for precise positioning. Horizontal accuracies between 4 and 6 cm RMS using ambiguity resolved PPP were confirmed for both fixed point and vehicle mounted receivers. This confirms that the QZSS L6 is an appropriate medium for broadcasting Australian made corrections as part of an effort to build a positioning infrastructure for the nation. As with past PPP research, the long convergence time remains the main issue with PPP solutions, in particular the GPS only solutions tested in the research will have periods of time in which bad satellite visibility or geometry will prevent ambiguity resolution. Numerous methods are being researched to achieve rapid convergence of PPP solutions. The efforts include partial ambiguity resolution, triple-frequency PPP and precise Ionospheric estimation. Next step in the research would be the generation and testing of precise Ionospheric corrections for Australia, although given the localized spatial correlation of the Ionospheric delays the QZSS L6 data rate may be insufficient to transmit the Ionosphere, in which case a complementary ground base delivery system would have to be considered. ACKNOWLEDGEMENTS This research is funded through the Australian Cooperative Research Centre for Spatial Information (CRCSI Project 1.11) and is a collaborative project between the CRCSI and the Japan Aerospace Exploration Agency (JAXA). The CRCSI research consortium consists of RMIT University, University of New South Wales, Victoria Department of Environment and Primary Industry, New South Wales Land and Property Information and Geoscience Australia. The authors would also like to thank JAXA for providing the GNSS and LEX receivers for testing. The effort of the CNES to provide real-time streams for PPP is gratefully acknowledged. REFERENCES Choy, S, Harima, K, Li, Y, Choudhury, M, Rizos, C, Wakabayashi, Y, Kogure, S (2015) GPS Precise Point Positioning with the Japanese Quasi-Zenith Satellite System LEX Augmentation Corrections. Journal of Navigation Collins, P, Bisnath, S, Lahaye, F, Heroux, P (2010) Undifferenced GPS Ambiguity Resolution Using the Decoupled Clock Model and Ambiguity Datum Fixing. Navigation, Journal of the Institute of Navigation 57

Ge, M, Gendt, G, Rothacher, M, Shi, C, Liu, J (2007) Resolution of GPS Carrier-Phase Ambiguities in Precise Point Positioning (PPP) with Daily Observations. Journal of Geodesy 82, 389-399. Geng, J (2010) Rapid Integer Ambiguity Resolution in GPS Precise Point Positioning. Ph.D. thesis, The University of Nottingham. Harima, K, Choy, S, Li, Y, Grinter, T, Choudhury, M, Rizos, C, Wakabayashi, Y, Kogure, S (2014) 'Performance of Real-Time Precise Point Positioning Using MADOCA-LEX Augmentation Messages, FIG Congress 2014.' Kuala Lumpur, Malaysia, 16-21 June. IGS RTS (2013) 'Real-time Service.' Available at http://rts.igs.org/ JAXA (2014) 'Interface Specifications for QZSS.' Available at http://qz-vision.jaxa.jp/use/isqzss/docs/is-qzss_16_rh_e.pdf Kishimoto M., Myojin E., Kogure S., Noda H., and Terada K. (2011), QZSS On Orbit Technical Verification Results, ION GNSS 2011, Portland, OR, 2011. Kallender-Umezu, P (2013) Enacting Japan's Basic Law for space activities: Revolution or evolution? Space Policy 29, 28-34. Lauricheese, D (2011) 'The CNES Real-time PPP with undifferenced integer ambiguity resolution demonstrator. The Institute of Navigation (ION) GNSS Conference (ION GNSS).' Oregon, Portland, USA, 20-23 September. Laurichesse, D, Mercier, F, Bertias, JP, Broca, P, Cerri, L (2009) Integer Ambiguity Resolution on Undifferenced GPS Phase Measurements and its Applications to PPP and Satellite Precise Orbit Determination. Navigation, Journal of the Institute of Navigation 56 Lauricheese, D (2015) Handling the Biases for Improved Triple-Frequency PPP Convergence. GPS World, Innovation column, April 2015. Rovira-Garcia, A., Juan, J.M., Sanz, J., Gonzalez-Casado, G. (2014), A Worldwide Ionospheric Model for Fast Precise Point Positioning, IEEE Transactions on Geoscience and Remote Sensing, Volume: 53, Issue: 8, pp:4596-4604 RTCM (2013), "RTCM STANDARD 10403.2 for differential GNSS," in RTCM 10403.2, ed. Arlington, Virginia: Radio Technical Commision for Maritime Services, 2013. Saito M., Sato Y., Miya M., Shima M., Omura Y., Takiguchi J., et al. (2011), "Centimeter-class Augmentation System Utilizing Quasi-Zenith Satellite," Proceedings of the 24th International Technical Meeting of The Satellite Division of the Institute of Navigation (ION GNSS 2011), Portland, OR, 2011. Suzuki T., Kubo N., and Takasu T. (2014), "Evaluation of Precise Point Positioning Using MADOCA-LEX via Quasi-Zenith Satellite System," ION International Technical Meeting 2014, San Diego, California, 2014. T. Takasu and A. Yasuda (2009), "Development of the low-cost RTK-GPS receiver with an open source program package RTKLIB," International Symposium on GPS/GNSS, pp. 4-6, 2009. Wang, J., and Y. Feng (2013). Reliability of partial ambiguity fixing with multiple GNSS constellations, Journal of Geodesy, Vol. 87, No. 1, pp. 1-14. Zumberge, JF, Watkins, MM, Webb, FH (1997) Characteristics and Applications of Precise GPS Clock Solutions Every 30 Seconds. Journal of Navigation 44, 449-456.