EE 573:

tentchoirΤεχνίτη Νοημοσύνη και Ρομποτική

15 Νοε 2013 (πριν από 3 χρόνια και 4 μήνες)

78 εμφανίσεις

EECE 621 Talk 7:

Autonomous

Swarm
-
Bot
Systems for WSNs

Prof. Sunggu Lee

EE Dept., POSTECH

Overview


This talk addresses general topics of mobility and use
of “swarm intelligence” techniques in WSNs


Mobility in WSNs


Introducing the possibility of mobile wireless sensor nodes increases
the range of issues, research problems, applications, etc., considered
with respect to WSNs


E.g., New problem of inserting
k

mobile sensors into a field of stationary
sensors in order to increase the sensing coverage to a specific desired level


Combines robot technology with sensor, wireless communication, and
microcomputer technology


There is a problem with energy conservation (mobile nodes will use a
lot more energy)


Applications may be limited, but such applications are still possible


In the future, miniature low
-
power robots may become feasible


thus, it
may still make sense to discuss mobility in WSNs as part of future
technology

2

Swarm Intelligence


Defn. and explanation of term


refer to wikipedia


The emergence of complex, global behavior from
interactions among many simple agents


Most commonly used examples


Bees in a beehive


Ants in an ant colony


An extremely interesting research topic, actively
studied in universities and other research institutions


Can be considered as a form of “biomimetics”


Also know as “biomimicry”


“… application of biological
methods

and systems found in
nature

to the study and design of
engineering

systems and
modern
technology
.”[wikipedia]


3

Implementation Considerations


Most hardware implementations of swarm intelligence
ideas use a small number of mobile nodes as part of
a “prototype system”


Due mainly to cost considerations


There may also be programming/control difficulties if
extremely large numbers of mobile nodes are used


Behavior of natural systems do not need to followed
exactly


Natural systems can serve as simply a guide or a source of
ideas regarding methods for solving certain problems


Engineering considerations may require changes to the basic
scheme

4

Swarm
-
Bot Systems [text]


Ch. 6 of [text] describes a project involving the use
of a small number of “ant
-
bots” used to demonstrate
cooperative intelligence concepts


Prototype system consists of a PCB board (with an
Atmega128L microcontroller and a CC2420 ZigBee
radio chip) attached to a programmable toy car


Refer to Figures 6.1 and 6.2 of [text]


Focus is not on the mechanical aspects of mobile sensor
nodes, but the design and implementation of communication
methods, positioning and cooperative behavior in such
systems

5

Cooperative Localization


Localization algorithm used here is different
from methods typically used in such systems
(to be discussed in next talk)


Coordinate systems used for robot systems


Absolute coordinates: each robot calculates its
coordinates with respect to an external common
coordinate system


Typically uses GPS or pre
-
established landmarks


Relative coordinates: each robot determines its
coordinates with respect to a reference point
within the operating environment

6

Relative Coordinate Localization


Initial phase


A selected coordinator moves around the circumference of a
circle and broadcasts a sequence of beacons


Each beacon packet contains coordinate of selected
coordinator, coordinator’s ID, and the transmitted power level


Each ant
-
bot within the transmission range of coordinator
can determine its own coordinates if it receives three
different position beacons


Based on triangulation or trilateration (refer to next talk)


Iterative phase


Suppose there exist “red” ant
-
bots (nodes that have not
received three separate position beacons)


Each ant
-
bot with known coordinates can repeat initial
phase with itself as the coordinator

7

Other Implementation Issues


Avoiding collisions among robots


Ant
-
bot uses the received signal strength indication (RSSI)
as main tool for avoiding collisions


RSSI can

also be used for distance measurements


Also uses information about the mutual orientations of ant
-
bots


Authors also claim to have developed a distributed approach
for determining dynamic routes for each ant
-
bot


Demo implementation


Text/graphics arrangement game


robots move to form
specific letters or shapes


Example: Refer to Figure 6.14 of [text]

8

Coordinator Election


A set of independently behaving nodes must send messages
back and forth in order to elect a unique coordinator node


Method used in this chapter


Sink node broadcasts Coordinator Election message


Each ant
-
bot backs off for a short period based on the RSSI (the
farther the distance, the shorter the backoff period)


If backoff time expires, ant
-
bot separately broadcasts a Coordinator
Election message


If Coordinator Election message is received before backoff timer
expires, the message is received and then a new backoff period based
on RSSI is

chosen


An ant
-
bot that receives two Coordinator Election messages
compute a backoff period based on RSSI (the farther …)


When backoff timer expires, ant
-
bot broadcasts Coordinator
message


If Coordinator message is received, an ACK Coordinator message is
sent and then a Stop Electing message is broadcast


9

10

[Figure 6.5 of [text]]

Position Measurement


If GPS is not available (low
-
cost or indoor
environment), then an indoor position
estimation method must be used


Most existing localization methods make use of
range measurements based on RSSI, time of
arrival (ToA) of a communication signal, time
difference of arrival (TDoA) or angle of arrival
(AoA)


RSSI is most commonly used because of its
simplicity


however, it is not very accurate


Distance estimate based on measurement of the strength
of received radio signal versus initial radio signal strength

11

Direction Measurement


Each ant
-
bot is equipped with an
electronic compass


The ant
-
bot knows which direction it is facing


Ant
-
bot can be programmed to turn in any
direction by giving desired angle and current
direction


Once target’s coordinates are known, each
ant
-
bot must plan a route to the target


Compute angle and distance of target


Equations given in Section 6.4.3 (pp. 175
-
176)

12

Minimap Integration


A fixed 8 x 8 minimap model is used to represent the
environmental data for an ant
-
bot in its coverage area


Ant
-
bot uses infrared or ultrasound sensors to determine positions
of obtacles


Based on approximate distance measurements to obstacles


Any time fixed parameter values are used (such as 8 x 8), the
selection of those parameter values must be justified, preferably
by analytical techniques


In this case, 8 x 8 minimap was used because of engineering
considerations


Wanted to use 64
-
bit (8 x 8) fixed packet sizes


Larger maps are created by combining the minimap
information provided by several adjacent ant
-
bots


Coordinator creates this combined map


e.g., Fig. 6.9 of [text]

13

14

[Fig. 6.9 of [text]]

Collaborative Path Planning


Proposed collaborative path planning
algorithm is a heuristic method based on
locations of obstacles and angle and
distance measurements


Actions of individual ant
-
bots must be controlled
determined such that collisions are avoided


Priority order of direction to be followed
determined based on assigned direction and
current location of ant
-
bot


Refer to Figure 6.10 of [text]


Refer to [text] for more details

15

16

[Fig. 6.10 of [text]]

17

[Fig. 6.11 of [text]]