110 likes | 297 Views
Acoustic Ranging in Resource-Constrained Sensor Networks. Branislav Kus ý, János Sallai, György Balogh, Miklós Maróti, Ákos Lédeczi { kusyb, sallai, bogyom, mmaroti, akos }@isis.vanderbilt.edu. Thursday, July 31, 2014. Wireless Sensor Networks. Inexpensive sensor nodes
E N D
Acoustic Ranging in Resource-Constrained Sensor Networks Branislav Kusý, János Sallai, György Balogh, Miklós Maróti, Ákos Lédeczi { kusyb, sallai, bogyom, mmaroti, akos }@isis.vanderbilt.edu Thursday, July 31, 2014
Wireless Sensor Networks • Inexpensive sensor nodes • Low-power microcontrollers • Small, integrated radio packages • Sensors (microphone, thermometer, magnetometer, light sensor, etc.) • Severe resource constraints (ram, processor speed, etc.) • Wide range of potential applications from Precision agriculture … … to military surveillance Thursday, July 31, 2014
Localization • Goal • Determine the geographical location of nodes of a sensor networks – crucial in many application scenarios • Requirements • Energy efficiency • Low cost • Resilience to noisy sensor readings • Localization commonly accomplished using range estimations between sensor nodes Thursday, July 31, 2014
Ranging techniques • Using RF • proximity based range estimation • RF signal strength information … not sufficiently precise range estimates • Using acoustic signals • Time of flight based • Time difference of arrival based … can do sub-centimeter precision, but not resilient to environmental disturbances • Ultrasonic ranging … more precise than audible sound, but has shorter effective range Thursday, July 31, 2014
Our approach • Acoustic ranging • Sample the acoustic signals • Digitally process the data • Reduce Gaussian noise with iterative sampling • Detect start of signal within the sampled window • Estimate distance from TOF • Target platform • Mica/Mica2 motes • 4 and 7.3 Mhz microcontrollers • 4 kB RAM • 38.4 kbit/s radio • Fixed frequency buzzer • Microphone, 17kHz sampling • Severe resource constraints Thursday, July 31, 2014
Measuring TOF • Signal source notifies the sensor via radio message that an acoustic signal is emitted • Sensor estimates TOF by measuring timebetweenreceiving the notification over the radio and detecting the start of the acoustic signal • Assume speed of sound is constant, d = c*TOF • Issues • Generating signal with sharp rising edge is infeasible on the available HW • Accurate detection of the start of a noisy signal is difficult Thursday, July 31, 2014
Signal at beacon d1 d2 d3 ls ls ls ls Signal at sensor Sampling intervals t0 t2 = t1+ls+d2 t3 = t2+ls+d3 t1 = t0+ls+d1 Addressing the Issues (1) • Computing the sample-wise sum of multiple sampled signals (each with the same expected signal form) • Gaussian noise will cancel out • “Multisampling” improves SNR by 10log(N), where N is the number of signals used Thursday, July 31, 2014
Addressing the Issues (2) • Digital filtering • Band-pass filter with integer coefficients • The ambient noise in the test recording was colored • matched filter was applied • Filter coefficients from the range [-4, 4] were acquired using a genetic algorithm • Calibration • Position of first peak in the filtered signal is detected (there may be further peaks caused by echoes) • Linear regression to map TOF measurements to distance Thursday, July 31, 2014
Results Thursday, July 31, 2014
Conclusion • Using simple signal processing techniques is viable for resource constrained sensor nodes. • Significant increase of precision and effective rage compared to other solutions on the same hardware. • Issues • Echoes cause false results • Effective range is limited • Power of the output signal • Size of memory buffer to hold samples in the sensor node Thursday, July 31, 2014
Questions Questions, Comments & Suggestions kusyb@isis.vanderbilt.edu sallai@isis.vanderbilt.edu Thursday, July 31, 2014