Design of an underwater acoustic pinger locator system. Student: Bradley Donnelly Supervisor: Associate Professor Karl Sammut Method The Plan: 4 hydrophones in an ultra-short baseline arrangement receive signals. Time differences between these signals are obtained by cross correlation. These measurements are used in the update stage of the particle filter to iteratively estimate the signal source. Introduction Acoustic pinger localization is a very important part of modern underwater technology especially autonomous vehicles. It can be used to help localise an autonomous vehicle in an underwater mission area. This is useful because electromagnetic signals do not penetrate the surface very far so traditional land based techniques such as GPS or radar become ineffective. It is for these reasons that AUVSI (Association for Unmanned Vehicle Systems International includes an acoustic pinger localisation task in many of their autonomous vehicle competitions The RobotX Maritime Challenge is an international competition consisting 15 teams from 5 countries . Each team is provided a WAM-V (Wave adaptive modular vessel) platform and tasked with designing a system of sensors, software and propulsion to turn the WAM-V into an autonomous vehicle. The WAM-V needs to be capable of performing various tasks autonomously. Such as the localization task which requires teams to search an area roughly 100m x 40m x 5m for a acoustic source. The source transmits is at a predetermined frequency between 25 – 40 kHz. Hardware: Teledyne Reson hydrophones , arranged so that spatial differences will occur in X,Y and Z dimensions, feed into a custom analogue board with adjustable gain and programmable filter. A Digilent Nexy4 FPGA board then performs the signal processing and adaptive features. With 4 sensors the time differences of arrival can be written as [3] System block diagram Cross correlation, given by , can be used to determine the time lag between two similar signals. This was implemented using a multiply accumulate block provided by Xilinx. To increase speed cross correlations were done in 64 sample batches. solving this for x, y and z is non-linear and quite complicated. The plane wave assumption can be used to significantly simplify the calculations this however breaks down over short distance. By implementing a particle filter an estimate can be simply derived without sacrificing accuracy. Multiply Accumulator used by the cross correlator Results Each component of this system was verified to work in either lab conditions or in simulation. Sadly there is currently no field data for this system as the majority of the data collection was indented to occur in Singapore during the competition. Whilst competing issues arose that caused delays. Due to these delays the possibility for field testing the system disappeared or obtain any raw data. I intend to rectify this over the coming weeks. Teledyne Benthos acoustic pinger Particle filter estimate after 1 iteration Particle filter estimate after 20 iterations (Zoomed in) K-Wave simulation environment highlighting TDOA Raw (simulated) hydrophone data Cross correlations from raw data show TDOA Each part of the hardware was verified, the digital pots communicate correctly via I2C and are capable of adjusting the gain of the op amp. The centre frequency and bandwidth of the programmable filter also change as expected. The FPGA implementation was also thoroughly tested using a combination of Vivado simulation and lab testing with the custom board. The algorithms themselves were predominately tested using a MATLAB acoustic simulation toolbox called K-wave. Flinders and AMC entry into RobotX [1] Example pinger localization course [2] Summary Acknowledgements This poster shows that each of the components required for acoustic pinger localization have either been verified to work individually under lab conditions or in simulation. The custom amplifier-filter board encountered various hiccups in development but in the end full functionality was achieved under lab conditions. The cross correlator that is implemented in fabric runs in real time and with appropriate functionality. The particle filter that was implemented in MATLAB worked very well. However this is not really an acceptable final result for my thesis , and needs to be seriously addressed, Over the next week I will get a full system test run and obtain field data for further testing. In terms of further work outside of thorough testing, looking into methods to overcome multipath issues. This would be helpful as several other teams at the competition claimed that they were struggling with this. Potentially doing some sort of initial chirp detection to find the original source and then relying on cross correlation to be unaffected by the reflections. [1] Photo taken by Jonathan Wheare [2] Maritime RobotX Challenge Preliminary Rules and Task Descriptions, 2014. Ver 2.6, 3 September. http://www.robotx.org/files/MRC-Rules-and-Tasks-2014-09-02.pdf [3] Cai, Z, Mohlenhoff, J, Puklavage, C. 2010. Acoustic Pinger Locator Subsystem. Senior Design, Group 19 3 May 2010. [4] Paull, L, Saeedi, S, Seto, M, Li, H. 2014. AUV Navigation and Localization: A Review. IEEE Journal of Oceanic Engineering, Vol 39, No 1, pp 131-149. Example multipath scenario [4]