A 3D ultrasonic positioning system with high accuracy for indoor application

Transcription

1 A 3D ultrasonic positioning system with high accuracy for indoor application Herbert F. Schweinzer, Gerhard F. Spitzer Vienna University of Technology, Institute of Electrical Measurements and Circuit Design, Gusshausstrasse 25/354, A-1040 Vienna, Austria Tel.: Fax: Abstract In the last years, a few Ultrasonic (US) positioning systems were presented which are primarily directed at indoor tracking of mobile devices. The US positioning system LOSNUS presented in this paper widens the application additionally supporting locating of static devices which can be numerously deployed communicating via a wired or wireless sensor network. However optimized for cheap realization, the system offers locating with high accuracy at medium high speed. The principle of operation, receiver design, and basic measurement results of reached accuracy are presented in the paper. Keywords: Indoor positioning system, device localization, ultrasonic ToF measurements 1. Introduction Indoor GPS systems which are able to localize mobile and static objects in buildings get increasing importance. Applications from different domains are well-known: tracking of persons or mobile devices, inspection of object locations, automatic device setup and security aspects in building networks [9]. Several systems based on US locating were presented in the last years which are based on time-of-flight (ToF) measurement using different design concepts. Basically, active tags attached with objects which have to be localized transmit US pulses which are received from a measurement system with at least three receivers fixed in each room. The start of transmission of the US signal is marked with an additional RF-signal which allows the receivers to measure the ToF between the start marker and the reception of the US signal. The tag position is calculated by a trilateration algorithm [12, 1, 3]. Some advantages show systems with reverse operation where the measurement system uses fixed transmitters and the tags receive the transmitter signals [5, 4]. In this case, security demands are fulfilled because tags are not sending by oneself and therefore cannot be detected by untrustworthy observers. Moreover, numerous tags can be localized in parallel. Applying a pseudo-trilateration algorithm, an explicit RF start marker is expendable. Instead of a ToF, timestamps of a local receiver clock are captured when receiving an US signal. The first captured timestamp is used as reference and differences of timestamps of other transmitters with the first one ( pseudo ToFs ) are used in computing the pseudo trilateration algorithm. At least three pseudo ToFs have to be provided if jointly using geometric information [6]. The reverse method and pseudo-trilateration are used by the system presented in this paper. 2. Principles of US positioning Some crucial points influence the applicability of US locating systems: locating speed (measurement rate respectively), effort of signal coding, and positioning accuracy. Some systems are working with uncoded US signals [7, 8, 12] which has the advantage that only narrow band US transducers are used being cheap and multiple available. However, this method is slow because a sufficient delay has to be included after each signal transmission 2-118

2 ensuring that the actual transmitted signal has reached all receivers and pendular echoes have decayed. Using coded signal transmission of several transmitters in parallel, the locating speed is maximized [3, 4] but the complexity of signal processing is significantly increased. Several signals with different amplitudes may overlap at the receiver which makes an effective signal separation necessary. The US positioning system LOSNUS (LOcalization of Sensor Nodes by UltraSound) presented in this paper uses a compromise: ToF measurements can be performed with high accuracy ( mm) by a unique chirp coded signal used by every fixed transmitter of the measurement system [2, 10, 11]. However, the transmitters are activated according to a time protocol which ensures that the transmitter signals are received in the original sequence. The transmitters are characterized by coding frequencies following the common chirp signal enabling the calculation of the receiver position based on the well-defined transmitter positions. Basically avoided signal overlapping has the advantage that one-bit quantization (only considering the sign) of the received signal which effectively simplifies signal processing delivers high quality results in the most receiver positions. 3. Measurement protocol of LOSNUS As with the most US positioning systems, minimum five or six transmitters are permanently mounted on the ceiling in each room. They are operated by a base station which provides differently coded signal frames for the transmitters. Activated sequentially, the transmitters can be supplied by demultiplexing of the output of a single amplifier. The time protocol of the measurement sequence is shown in Fig. 1. The locating cycle is started by a synchronization burst signal with constant frequency performed by several transmitters. This frequency is determined by a band-pass filter of the receiver. After a short intermission, the measurement cycle continues with the transmitter identifying frames. Fig. 1. Sequence of signal transmissions ensuring device synchronization and non-overlapping reception of transmitter identifying frames. Fig. 2. Sequence of signal transmissions ensuring device synchronization and non-overlapping reception of transmitter identifying frames. Transmitted frames start with a lead-in part of constant frequency which allows the decay of transient effects of the transducer (refer to Fig. 2). At least three periods of about 30 khz are used for this task. The following part is a chirp signal enabling high resolution and precise 2-119

3 distance measurement. The chirp with a length of 256 s is identical for each transmitter which allows calculating the pseudo ToFs from all transmitters with the same reference signal. The frequency of the chirp is increasing in a range of about 30 khz to 70 khz. To be able to include the fixed transmitter position in the trilateration algorithm, the chirp part is followed by one or several parts of transmitter identification. To enable continuous frequency changes, the chirp and the constant coding frequency parts of defined length (100 s) are connected with changeover phases of 100 s as illustrated in Fig. 2. Both, coding frequencies and changeover phases are characterizing the transmitter and can be evaluated in different ways. Using three coding parts in the frame as done in the test system, a number of minimum 512 different codes can easily be realized which is in many cases enough for being uniquely assigned to all transmitters of a building. In this case, the total frame length is less than 1 ms. Both, time protocol and high measurement accuracy deliver significantly improved system properties compared with other systems. A test system was installed with 6 fixed transmitters on a laboratory wall allowing position measurement rates up to 10 measurements per second. Computing a pseudo-trilateration algorithm, the high accuracy of ToF measurements result in a low uncertainty of the device position depending on geometric constellation and used algorithm. 4. US receiver of localized device The locating method is especially designed to allow simple receiver constructions. A compact sensor system with microcontroller, US receiver, and wireless network hardware was built to display the use of ultrasonic positioning in applications of arrays of static sensor devices: the device position is a crucial parameter for processing sensor data of parallel sensors and further enables advanced methods of automatic integration of the device in the network (see Fig. 3). Fig. 3. Compact sensor system including US receiver, microcontroller, and wireless network access. The US receiver hardware allows to extract transmitter frames from the continuously received signal. The US signal is received by a small broadband electret microphone. After amplifying and bandpass filtering to avoid audio and high frequency disturbances, the signal is one-bit quantized with 1 MSample/s. When a signal trigger with a threshold over the noise is activated, a timestamp is captured and storing of data is synchronously started to hold a frame of about 1kbit. The microcontroller checks the data sequence in parallel whether containing lead in or other characteristics of a frame. If checking is positive, the data block is transmitted and otherwise discarded. This method reduces the amount of data to be wirelessly transferred after a locating sequence to typical 6 to 10 kbit. 5. Signal processing and measurement results The transferred data blocks are received by a central computer which evaluates the device position. At first, the data blocks are analyzed by correlative signal processing to get the chirp position in the frame. Together with timestamp and constant sampling rate, the chirp position 2-120

4 adds up to the reception time with the resolution of the sampling rate. Further, the transmitter coding is analyzed. Data blocks without unambiguous information are discarded. Fig. 4. Received frames computed by correlative signal processing. Frames 4 and 8 are disturbances. The frames are independently evaluated and depicted in the figure at the appropriate points of time of chirp positions relative to the first frame. The transmitter codes 1 to 6 are written above. Transmitter 2* shows a pendular echo after the first reception of a signal of transmitter 2. The numerical results of this locating sequence are given in Table 1. Figure 4 shows the correlative chirp analysis of a typical positioning sequence with 8 transmitted data blocks which are depicted at their reception times. 6 transmitters and 2 disturbances are recognized. The frame of transmitter 2 is enlarged because it contains a second frame of the same transmitter which is eliminated. Table 1 shows characteristics of these 8 data blocks. For application of pseudo-trilateration, differences between transmitter timestamps and the reference timestamp (transmitter 1) are computed and corrected by the frame delays listed in table 2 which delivers the pseudo ToFs of transmitter 2 to 6. Table 1. Received frames of a typical measurement sequence. The pseudo ToF ([ms]) is the difference between transmitter timestamps #2 to #6 and #1 ([μs]) corrected by the delayed transmitter start times (see Table 2). Frame # Transmitter Time- Difference rela- Pseudo Chirp Correlation Transmitter Coding # stamp tive to #1 [ms] ToF Max. 256 % Max. 599 % Table 2. Transmitter start times and room coordinates. The transmitters are mounted on a wall; the base of the system of coordinates is at the bottom left corner; x is directed horizontal right and z vertical upwards. Transmitter # Start of frame [ms] Room coordinate x Room coordinate y Room coordinate z m m m m m m m m m m m 2.90 m m m m m m m 2-121

5 A measurement series of 35 positioning sequences with the same receiver position as in table 1 was analyzed. The mean values of the pseudo ToFs of transmitter 2 to 6 and the corresponding distances are evaluated together with the empiric standard deviation (Table 3). The pseudo-trilateration algorithm delivers different results depending on the geometric situation. Four differences are used in a simple linear algorithm which has reduced accuracy, refer to [6]. An approximation algorithm only using three differences would deliver better results. Getting five differences enables using a multilateration which calculates the device position with significantly increased accuracy. Table 3. Measurement series of 35 locating sequences. Receiver in position (1.73 m, -3.2 m, 1.1 m); speed of sound 343 m/s; trilateration result (x, y, z) in that line which transmitter is excluded. Transmitter # Mean value of pseudo ToFs [μs] Mean value of difference of distances [mm] Empiric standard deviation [mm] Trilateration result of 4 differences excluding 1 transmitter x [m] y [m] z [m] 2 945, , , , , Multilateration result of 5 differences, standing right Conclusion A 3D device positioning system based on ultrasonic measurements was presented which shows significantly improved accuracy and measurement rate. It allows simple and compact constructions of devices to be localized which is beside of mobile devices also interesting for static devices, e.g. sensor devices. A test system was set up and basic measurement results were presented in the paper. References 1. M. Addlesee, et al. Implementing a sentient computing system. IEEE. 2001, 34(8). 2. H. Elmer, H. Schweinzer. High resolution ultrasonic distance measurement in air using coded signals. IEEE IMTC/2002, Anchorage, USA M. Hazas, A. Ward. A novel broadband ultrasonic location system. UbiComp 2002: Ubiquitous Computing, Göteborg, Sweden M. Hazas, A. Ward. A high performance privacy-oriented location system. PERCOM 04: Pervasive Computing and Communications M. Langheinrich. Privacy by design principles of privacyaware ubiquitous systems. UbiComp2001: Ubiquitous Computing, Atlanta, USA A. Nishitani, Y. Nishida, T. Hori, H. Mizoguchi. Portable Ultrasonic 3D Tag System Based on a Quick Calibration Method. IEEE, Netherlands N.B. Priyantha, et al. The Cricket Location-Support System. ACM MOBICOM C. Randell, H. Muller. Low cost indoor positioning system. UbiComp, USA H. Schweinzer, W. Kastner. Systems with numerous low-cost sensors new tasks and demands for fieldbusses. 5 th IFAC Int. Conference FeT 03, Aveiro, Portugal H. Schweinzer, H. Elmer. High resolution ultrasonic distance measurement systems using pulse compression and their application. ISMTII 03, Hongkong H. Schweinzer, P. Krammer. A 3D-Location System Optimized for Simple Static and Mobile Devices. 6 th IFAC Int. Conference FeT 05, Puebla, Mexico A. Ward, A. Jones, A. Hopper. A New Location Technique for the Active Office. IEEE Personal Communications Magazine. 1997, vol. 4, No