Upload
others
View
6
Download
0
Embed Size (px)
Citation preview
Monash University AUSTRALIA 1A Real Time DSP Sonar Echo Processor - IROS’2000
A Real Time DSP Sonar EchoProcessor#
by
Andrew Heale and Lindsay Kleeman
Intelligent Robotics Research Centre
Department of Electrical and Computer Systems Engineering
Monash University, Victoria, AUSTRALIA
www.ecse.monash.edu.au/centres/IRRC
# Funded by an Australian Research Council Large Grant
Monash University AUSTRALIA 2A Real Time DSP Sonar Echo Processor - IROS’2000
Contents
u Introduction
u Processing
u Results
u Conclusions
Monash University AUSTRALIA 3A Real Time DSP Sonar Echo Processor - IROS’2000
Introduction
u Robot systems are often sensor bound:– eg map building, localisation, obstacle avoidance
are limited by sensors, not algorithms
– require reliable, fast, accurate cheap sensing.
u Sonar has been seen as unreliable, inaccurate.– eg Polaroid ranging module is a poor angle sensor
u DSPs allow sonar echo real time processingwith more accurate range/bearing than laserrange finders.
Monash University AUSTRALIA 4A Real Time DSP Sonar Echo Processor - IROS’2000
Introduction (continued)
u This paper introduces a new DSP sonarsensor that achieves optimal signalprocessing at near real time rates.
u Sensor reports range and bearing to targetsto 5.4 m at 27 Hz.– Speed of sound alone imposes a limit of 30 Hz.
u Bearing and ranges errors are dominated byair conditions - typical still air gives:– 0.1 degrees, 0.2 mm error standard deviations.
Monash University AUSTRALIA 5A Real Time DSP Sonar Echo Processor - IROS’2000
DSP Sonar System
DSP Sonar system
T T
R R
Werrimbi
Monash University AUSTRALIA 6A Real Time DSP Sonar Echo Processor - IROS’2000
Custom Built Receiver
u 1MHzsampling on2 channels
u All in one box– reduces noise levels
u Fast processing by DSP
33MHz AD2181 DSP
+ ROM
LPF + 12 bit ADCs
High speed buffered UART
300 V FETTransmitter
Drive Circuit
Gain
PC
T
R1
R2
Monash University AUSTRALIA 7A Real Time DSP Sonar Echo Processor - IROS’2000
Overview of DSP Processing
u On the fly pulse extraction
u Template matching– parabolic interpolation for sub sample
estimation
u Correspondence
u Triangulation
Monash University AUSTRALIA 8A Real Time DSP Sonar Echo Processor - IROS’2000
Example of a Receiver Signal
Monash University AUSTRALIA 9A Real Time DSP Sonar Echo Processor - IROS’2000
Apply threshold = multiple of standarddeviation of noise.
Monash University AUSTRALIA 10A Real Time DSP Sonar Echo Processor - IROS’2000
Remove inter-pulse signal below thresholdfor at least 30 samples before and after pulses.
Monash University AUSTRALIA 11A Real Time DSP Sonar Echo Processor - IROS’2000
Summary of thresholding with pre andpost triggering
Monash University AUSTRALIA 12A Real Time DSP Sonar Echo Processor - IROS’2000
Pulse Splitting - based on signal peaksthat are clear of surrounding peaks.
Monash University AUSTRALIA 13A Real Time DSP Sonar Echo Processor - IROS’2000
Template Matching
u Templates are pre-computed echo pulseshapes used in the arrival time estimation.
u Shape depends on arrival angle and range.
u This dependency has been accuratelymodelled - see [Kleeman&Kuc IJRR 1995]
u Thus the template set can be generated froma measured echo at normal incidence at 1m.
Monash University AUSTRALIA 14A Real Time DSP Sonar Echo Processor - IROS’2000
Template shape varies withangle to transducer
0 deg 2 deg 4 deg6 deg
8 deg 10 deg 12 deg 14 deg
Transducer
Monash University AUSTRALIA 15A Real Time DSP Sonar Echo Processor - IROS’2000
Templates shifted to best match the pulses using correlationof the template with the pulse. Shape (not amplitude) are matched by a correlation.
Template Matching
Monash University AUSTRALIA 16A Real Time DSP Sonar Echo Processor - IROS’2000
Monash University AUSTRALIA 17A Real Time DSP Sonar Echo Processor - IROS’2000
Receiver Data Association
u Left and right receiver arrival times areassociated based on:– arrival times consistent with small receiver
spacing
– amplitudes matched
– correlation coefficients > 95%
Monash University AUSTRALIA 18A Real Time DSP Sonar Echo Processor - IROS’2000
Door jamb experiment
Reported sensor readings
Monash University AUSTRALIA 19A Real Time DSP Sonar Echo Processor - IROS’2000
MappingExperiment Sonar features from
moving sensor
Mouldings
RobotPositions
Wall
Mouldings
Door jamb
Monash University AUSTRALIA 20A Real Time DSP Sonar Echo Processor - IROS’2000
Fused Targets
Raw target data in closeproximity is fused basedon a weighted average.
Weights are determinedby amplitude andcorrelation values.
Monash University AUSTRALIA 21A Real Time DSP Sonar Echo Processor - IROS’2000
Conclusions and Future Work
u DSP sonar implementation allows:– local processing obviating high speed data
communications
– real time processing is achieved
– portability.
u Futher work:– real time on the fly target classification.
– interference rejection with double pulse coding(presented ACRA 2000, Melbourne Australia).
– Real time SLAM.