1. Self–localization of Wireless Sensor Nodes by means of Autonomous Mobile Robots A note on the use of these ppt slides: We’re making these slides freely available to all, hoping they might be of use for researchers and/or students. They’re in PowerPoint form so you can add, modify, and delete slides (including this one) and slide content to suit your needs. In return for use, we only ask the following: If you use these slides (e.g., in a class, presentations, talks and so on) in substantially unaltered form, that you mention their source. If you post any slides in substantially unaltered form on a www site, that you note that they are adapted from (or perhaps identical to) our slides, and put a link to the authors webpage: www.dei.unipd.it/~zanella Thanks and enjoy! A note on the use of these ppt slides: We’re making these slides freely available to all, hoping they might be of use for researchers and/or students. They’re in PowerPoint form so you can add, modify, and delete slides (including this one) and slide content to suit your needs. In return for use, we only ask the following: If you use these slides (e.g., in a class, presentations, talks and so on) in substantially unaltered form, that you mention their source. If you post any slides in substantially unaltered form on a www site, that you note that they are adapted from (or perhaps identical to) our slides, and put a link to the authors webpage: www.dei.unipd.it/~zanella Thanks and enjoy!

2. Department of Information Engineering – University of Padova – Italy Self–localization of Wireless Sensor Nodes by means of Autonomous Mobile Robots Andrea Zanella, Emanuele Menegatti and Luca Lazzaretto 2007 Tyrrhenian Workshop on Digital Communications TIWDC 2007 This work was supported by the University Project “RAMSES2: integRation of Autonomous Mobile robots and wireless SEnsor networks for Surveillance and reScue”

3. 12 September 2007 3 Andrea Zanella WSN vs AMR Pros –Low cost (hundreds of devices)‏ Cons –Limited computing capabilities –Limited memory –Limited energy capacity –Limited transmission range/speed –No or very limited mobility Cons –High cost (a few devices)‏ Pros –High computational capabilities –Large memory –Large energy capacity –Large transmission range/speed –Advanced mobility WSN: Wireless Sensor NetworkAMR: Autonomous Mobile Robot RAMSES 2

4. 12 September 2007 4 Andrea Zanella RAMSES 2 RAMSES 2 : integRation of Autonomous Mobile robots and wireless SEnsor networks for Surveillance and reScue WSN monitors strategic areas –earthquakes, fire, land/snow-slide, chemical hazards,... In case of danger, AMRs team is activated –WSN rise a fire alarm? AMR squad is driven by the WSN to the hot area AMR with fire extinguishers cooperate to extinguish the fire Other AMRs establish ad hoc multihop connection to stream high– quality video to a control centre

5. 12 September 2007 5 Andrea Zanella AMR  WSN: AMR–aided WSN maintenance AMR can work as a data mule, collecting data from nearby nodes and, then, releasing them in another location, perhaps over connectivity holes –Improve WSN connectivity –Alleviate energy consumption –Increase data reliability AMR can be used to place new nodes in the WSN where needed A single AMR equipped with sophisticated and reliable transducers can be used to calibrate the cheap transducers of WSN nodes AMR can be used to improve self-localization of WSN nodes

6. 12 September 2007 6 Andrea Zanella Self-localization Problem statement: –A bunch of sensor nodes are hand-placed in a given room –Each node needs to localize itself with respect to a common reference system –Nodes are only equipped with RSSI transducer –Localization error shall be reduced as much as possible –WSN nodes shall dissipate as few energy as possible State of the art: –Plenty of localization algorithms in literature –Range-free Not require ranging capabilities Good with dense networks and not very harsh propagation environments Poor performance in indoor and low density WSN

7. 12 September 2007 7 Andrea Zanella Range-based self localization A few beacon (anchor) nodes are placed in know positions in the area Beacons periodically broadcast their own positions Other nodes try to estimate their distance from beacons and infer their own position on the area by using different methods

8. 12 September 2007 8 Andrea Zanella Range-based self localization: issues Range-based localization problems –Very sensitive to ranging errors Channel characteristics (shadowing, multipath, asymmetry,...)‏ –Very sensitive to loose calibration Nodes with identical setting may reveal differences in transmission power or reception sensitivity Localization algorithms leveraging on cooperation among different nodes usually neglect such calibration misalignments –Extra hardware required for good performance (ultrasounds transceivers, multiple antennae, several beacons)‏ high costs reduced flexibility –Very poor performance in indoor and low density WSN

9. 12 September 2007 9 Andrea Zanella AMR–aided WSN Self-localization AMR can alleviate many of such problems! How does it work? –AMR moves in the room and estimates its own position using on- board motion sensors (odometers)‏ –Periodically AMR broadcasts its current positions Then? –The number of (virtual) beacons can be indefinitely increased –Transmissions are performed by a single device, then calibration issues are mitigated –AMR self-estimates its own position, then handy beacons placement is avoided –AMR might support expensive equipments since they have not to be replicated in several devices

10. 12 September 2007 10 Andrea Zanella Experimental Set up EyesIFX sensor nodes –Infineon Technologies. –19.2 kbps bit rate @ 868 MHz –Light, temperature, RSSI sensors SIGNET IAS AMR Bender –self-made, based on Pioneer 2 ActivMedia platform –Linux OS with Miro middleware –ATX motherboard –1,6 GHz Intel Pentium 4, 256 MB RAM, 160 GB HD EyesIFX connected to ATX via USB + EyesService class added to Miro –Omnidirectional camera, odometers

11. 12 September 2007 11 Andrea Zanella Measurements setting Empty corridor of 4.5m × 10m (ceiling 4m). Robot moves along three parallel lines @ average speed 240 cm/s Robot coordinates broadcasted every 50ms through the on-board EyesIFX node Ten static sensor nodes form an incomplete lattice Nodes receive messages & store coordinates & RSSI

12. 12 September 2007 12 Andrea Zanella Channel model Path loss channel model: received power P i @ distance d i Received power Transmitted power Path loss coefficient reference distance environmental constant real transmitter- receiver distance Shadowing fast fading

13. 12 September 2007 13 Andrea Zanella How harsh is the indoor radio channel? Random variations due to shadowing and fading obscure the log-decreasing law for the received power vs distance RSSI based ranging is VERY noisy!

14. 12 September 2007 14 Andrea Zanella Channel model parameters fitting Low-pass filtering data we can extract the underlying log law The best fitting of the filtered measures with the theoretical relation gives P Tx + K = −30.5 dBm  = 1.5 d 0 = 10 cm

15. 12 September 2007 15 Andrea Zanella Shadowing distribution QQ.-plot shows that shadowing (in dB) is approximately Gaussian with μ=−0.0348 ± 0.0860 dB  = 6.339 ± 0.0614 dB (95% confidence interval)‏

16. 12 September 2007 16 Andrea Zanella Experimental results: multilateration Multilateration is a range-based localization algorithm that offers –very basic calculations (low complexity) –limited memory occupancy –no need for node transmissions  reduce interference & energy cost In theory –For each received message, nodes trace a circle centered on the beacon and having radius equal to the estimated distances from the beacon –Ideally, circles intersect in a single point on a surface which gives the node location In practice –Nodes have limited computation capabilities  area is divided in cells –Channel impairments require to consider rings instead of circles –For each received message, nodes increased by one the weight of the cells covered by the ring centered on the beacon and having radius equal to the estimated distances from the beacon –The node us located within the cells that scores the maximum weight

17. 12 September 2007 17 Andrea Zanella Results: localization with N virtual beacons Multilateration on RSSI samples randomly picked from the full data set Conversely to what expected, more samples do not improve localization!

18. 12 September 2007 18 Andrea Zanella Why taking highest RSSI? Noise free RSSI RSSI + =RSSI + |  | RSSI - =RSSI - |  | d+d+ d-d-

19. 12 September 2007 19 Andrea Zanella Results: localization with the “highest” RSSI Localizing over sorted RSSI yields much better performance –Ranging errors are more relevant when considering lower RSSI values due to the logarithmic nature of the path loss model Localization improves as the number of samples reduces!

20. 12 September 2007 20 Andrea Zanella Conclusions RSSI-based localization show very poor performance in indoor environments –Shadowing, fading, calibration errors,... Although AMR can alleviate some of the primary causes of localization errors, standard localization techniques still yield poor performance in indoor environment! Nevertheless, the possibility of drastically enlarging the number of collected samples and the greater computational capabilities of AMRs permit to define more performing algorithms!

21. 12 September 2007 21 Andrea Zanella Discussion Question time! Thanks for the attention