Speech Transcript - BIO-INSPIRED SWARMING MODELS FOR DECENTRALIZED RADIO ACCESS INCORPORATING RANDOM LINKS AND QUANTIZED COMMUNICATIONS

0:00:15	so i think it's actually much for so what would like to say just so yeah the proposed a new
0:00:19	and for the position of
0:00:21	a special session makes a pleasure for me to be here
0:00:24	oh today will talk about a but it's at the more is what most applied to this to distributed resource
0:00:30	allocation problem in can be to systems
0:00:32	where the communication or more the second users
0:00:35	are supposed to be a a a a a a but randomly but i
0:00:41	a particular we will talk about social foraging
0:00:44	social sort
0:00:46	these systems are typically made up of a relation simple agents
0:00:49	in a a lot of locally to each other
0:00:52	and that with the empire
0:00:54	agents
0:00:54	follow a very simple rule
0:00:56	and even if there is not centralised four
0:00:59	need for um and then leads to
0:01:01	the merchants of one halogen original here that usually is unknown to the individual
0:01:07	that are examples of course clue don't call is
0:01:09	bird flocking willing
0:01:13	the application of which we want
0:01:15	to translate
0:01:16	five is part more
0:01:18	he's cognitive radio
0:01:19	we know how it this case
0:01:21	we have
0:01:22	to to kind of the users
0:01:23	primary users
0:01:25	that
0:01:26	um
0:01:27	are the owners
0:01:28	the red
0:01:29	and the second are you
0:01:31	these are kind of the users
0:01:33	that's a sense the channel in an opportunistic way
0:01:36	and trust me you local by channel
0:01:38	oh have
0:01:39	a second data the network used to get a dog
0:01:41	so we have start
0:01:43	without a central node
0:01:44	that that's controlled the access to the medium
0:01:47	that
0:01:47	our problem is how to coordinate yeah
0:01:50	from what actually many second the users
0:01:52	yeah know like this central i
0:01:54	a i just i mean my what condition
0:01:56	then
0:01:57	hence
0:01:57	in this context the by some
0:01:59	simple but robust
0:02:01	a a fraction mechanism
0:02:03	can be all real in
0:02:06	in particular we propose
0:02:08	at the strip of the rest are location based on board
0:02:10	they they that nearby by agent
0:02:12	and track beat each other
0:02:14	and for so um presents a location then prevents correlations among the location of different users
0:02:20	and a is crazy
0:02:22	so the search of the most of robert's slot to be used
0:02:25	can be modelled as the motion
0:02:26	or was formal press source
0:02:28	the rest of domain
0:02:30	yeah
0:02:30	i
0:02:31	i frequency plane or the frequency a
0:02:33	looking for for edge
0:02:35	for a represents a actually a function inversely proportional
0:02:38	the if you know level
0:02:40	sense by the
0:02:42	and that's work would tend to be in this time frequency region where the that's interference
0:02:47	what's that is fine to be one
0:02:49	i always regions
0:02:51	in the location of different users
0:02:52	and you might spread
0:02:53	in rest
0:02:56	rest of the location can be more the that as the this for like mean they of this sort of
0:03:01	for word
0:03:03	in particular
0:03:05	i
0:03:05	the knows the wrestlers that
0:03:07	so that was issue
0:03:08	sure sure of chosen by note i for is red
0:03:11	in the rest of me
0:03:14	can be in this case the entries of is back can get time slot the frequency
0:03:17	chump
0:03:19	second i i
0:03:20	is the interference profile sense
0:03:23	on the rest
0:03:24	but may i know i
0:03:25	so in this case the E power force there so uh
0:03:29	a a firm that want to me my
0:03:31	J A and J R
0:03:33	represents the a function are also to
0:03:35	they allowed to implement
0:03:38	a features of a fraction of also
0:03:40	between the
0:03:41	the the rest of
0:03:43	yeah J
0:03:44	are the entries of an identity matrix of the graph
0:03:47	this described
0:03:48	special the hall
0:03:50	oh
0:03:50	now
0:03:51	in particular
0:03:52	it is good are what one word zero
0:03:55	if no G is
0:03:58	you is a need of one of all not that so if it is inside scoring rate
0:04:02	well wise
0:04:03	so it's in this case what one
0:04:05	yeah yeah the wife
0:04:08	it's important remark
0:04:09	that's even a this is warming up guns
0:04:12	you know issue
0:04:13	so the secondary users are
0:04:15	i'm not moving at to just the rest
0:04:17	are moving in this rest to me
0:04:19	these with visions a J
0:04:21	the best this a supposition
0:04:23	of
0:04:24	so they are related to this special edition
0:04:29	attraction and repulsion can be chosen instead several
0:04:32	weight however allow for a simple mathematical
0:04:35	they should be
0:04:36	uh we have chosen a um
0:04:38	um
0:04:39	what for i
0:04:39	from
0:04:40	function
0:04:41	and that is financial bound
0:04:42	vol
0:04:43	so that's
0:04:44	the a joint you here
0:04:45	okay a with respect to these
0:04:47	in the rest of this assume this kind of four
0:04:49	it's important to note is that there is a unique
0:04:52	i
0:04:52	of the a fraction propulsion
0:04:54	potential actual function
0:04:56	these
0:04:56	mean
0:04:57	is really
0:04:58	to the so called it really this done
0:05:00	in the biological literature
0:05:02	is the distance of which a fraction
0:05:04	and also a
0:05:05	and can be set
0:05:07	choosing rubber is good vision
0:05:09	C and C a
0:05:10	yeah a action at of pops concert
0:05:12	what
0:05:13	in in our setting
0:05:14	there are also is useful to a hold collision
0:05:17	between in the location of different used
0:05:19	so we choose the i believe in this
0:05:21	we proportional the
0:05:23	this that
0:05:24	in the me
0:05:26	that can be and
0:05:27	of the
0:05:29	but
0:05:29	channel
0:05:30	for for for example the duration
0:05:31	of the time and times
0:05:33	so a the right solution to the previous problem can get you the
0:05:37	for a a simple steepest simple approach
0:05:40	where each no is at the position of its own wrestlers
0:05:44	and the rest
0:05:45	really according to the agree
0:05:47	gender the
0:05:49	fourteen potential function
0:05:50	i shows
0:05:51	four
0:05:52	it's important to remark
0:05:53	that these
0:05:54	rule
0:05:56	to be he needs
0:05:57	the computational
0:05:58	just not wrong
0:05:59	so a lot of gradient of this
0:06:02	it's five
0:06:03	a to the point X i
0:06:04	it well
0:06:05	which we are also
0:06:07	or some force
0:06:08	that the rest of six i are receiving
0:06:11	a a from the results of the need
0:06:14	J
0:06:15	it's was important from that's at the E
0:06:18	each of that
0:06:18	row
0:06:19	information to each with neighbour
0:06:22	a a a a a a the information about the position he's
0:06:25	he's going to i
0:06:28	this is
0:06:29	whole
0:06:30	they
0:06:30	read red provide represents it
0:06:32	a position of the psd L
0:06:35	a primary user
0:06:37	and and that is changing
0:06:39	time
0:06:39	simulation
0:06:40	the we though that are moving on the
0:06:43	frequency a
0:06:44	represents this
0:06:45	a a a a a uh uh this
0:06:46	the
0:06:47	i i i i and the red it is pressed
0:06:49	that are mean trying to find
0:06:51	a position where there is more for so less interference
0:06:55	a at the same time
0:06:56	so is fine
0:06:57	a to require on
0:06:59	a is what we see that so it's
0:07:01	they are trying to keep
0:07:02	this warm of a spread a small as possible and the would lead someone allocation
0:07:10	but
0:07:11	i'm going to present one of the main result
0:07:13	you one of the more
0:07:15	you features of this or with that is
0:07:18	intrinsic of of implementing of the centralized make
0:07:21	for sparse
0:07:22	chuck
0:07:24	i said before that
0:07:26	basically is movement movement happens
0:07:28	in um
0:07:29	the in the rest of me so it's kind of a big issues
0:07:32	but the communication
0:07:34	between
0:07:35	no
0:07:36	for some
0:07:37	in this part
0:07:38	so
0:07:39	basically
0:07:40	to nodes that that actually are the course for each other
0:07:43	can exchange information and can hold
0:07:46	to use the same channel
0:07:48	that
0:07:49	to to note that that actually a far away from each other
0:07:54	is no
0:07:54	is not
0:07:55	actually
0:07:56	they don't communicate with other
0:07:58	so they don't perceive repulsion
0:08:00	and they are allowed to use
0:08:01	same channel use the same press
0:08:03	a so
0:08:04	but
0:08:06	or in a a a a in this example we particular we're
0:08:09	to a network composed of a hundred nodes
0:08:12	and with you there in just a a weighted channel of a these node so
0:08:16	share
0:08:17	with different colours we identified a different channels they cannot board
0:08:22	where yeah
0:08:23	and that's it
0:08:24	as we can see it never happens choosing probably you can these
0:08:28	we
0:08:29	yeah
0:08:30	the rest
0:08:31	that's two nodes
0:08:32	are are in this say
0:08:34	a a rest
0:08:35	but a if they are far such as far away
0:08:38	and they don't have that i
0:08:43	that's now everything on the time
0:08:46	this is just a a
0:08:48	and easy down
0:08:49	where we can peaks
0:08:50	the position of the primary
0:08:52	i
0:08:53	and the money that
0:08:54	he
0:08:54	where we can is are position
0:08:56	to use
0:08:57	is
0:08:58	see
0:08:59	in the middle right dots
0:09:01	and uh uh we can see out is warm in to people's in for each other
0:09:06	feel the gas
0:09:07	let
0:09:07	uh a the primary users
0:09:09	and that that same time
0:09:11	you might spread and would lead you
0:09:13	these
0:09:15	each other
0:09:16	however is is
0:09:17	is is how well knowledge
0:09:19	primary users are you
0:09:21	four and a
0:09:22	at least i the least
0:09:24	so what we propose
0:09:26	it's
0:09:26	a model of the primary users so
0:09:28	if
0:09:28	time
0:09:30	probably in some knowledge about
0:09:32	the it is some is unknown
0:09:34	right
0:09:36	so in particular we propose
0:09:38	um that the primary users T V
0:09:40	when
0:09:41	channel
0:09:42	can be more as oliver james and i'm
0:09:45	i the right by the transition rate
0:09:48	from i a lot
0:09:49	um
0:09:50	from i one you
0:09:52	now suppose the that's this
0:09:54	patients are no
0:09:55	really efficient
0:09:57	and it is also known the average power yeah i
0:10:00	all one over a some channel
0:10:02	it is possible to about
0:10:03	close of four
0:10:04	the
0:10:05	temporal wrote it of the expect
0:10:08	you
0:10:08	condition it to the observation time what C
0:10:11	so i i i you particular are you the channel and T what is your sense of i
0:10:16	this is the expression
0:10:17	of them from either
0:10:20	otherwise
0:10:21	we have that
0:10:22	what is a portal
0:10:23	use a port and because they you they allow house
0:10:26	the construction the construction of a my
0:10:29	all
0:10:30	expected interference
0:10:32	where this warm actually pen
0:10:34	no okay
0:10:35	for time frequency slots where there
0:10:38	that's that's interference
0:10:40	let that
0:10:41	one
0:10:42	such is fine
0:10:43	the basic required
0:10:44	about
0:10:45	a receive S
0:10:47	and
0:10:48	is an example actually
0:10:50	and
0:10:51	in in the the expected interfere
0:10:54	time
0:10:54	it's use
0:10:55	right
0:10:56	right
0:10:57	using a grey scale
0:10:58	so it you can of the time
0:11:00	here is
0:11:01	i think people to zero
0:11:03	we have to these joint
0:11:05	um
0:11:06	yeah yeah easy
0:11:08	is it that
0:11:09	a right by the right
0:11:11	three his joint
0:11:12	uh uh actually um
0:11:14	i
0:11:15	or for by the dark
0:11:17	area so there areas
0:11:18	where it is more
0:11:20	right
0:11:20	more
0:11:21	and
0:11:22	as as we can see this work
0:11:24	from the initial position
0:11:26	the fine read location
0:11:27	and so the rest of this fine i a location
0:11:30	close that was in in time
0:11:32	where the prediction of the variance
0:11:34	so is better
0:11:35	on one i usually at of spectrum
0:11:37	what is be satisfy
0:11:40	force
0:11:40	it it required
0:11:41	about
0:11:43	to each
0:11:43	yeah
0:11:46	in particular at this whole we fall i i four was
0:11:49	all the fact that really stick channel
0:11:52	i
0:11:53	have
0:11:53	on this
0:11:54	or procedure
0:11:55	so that that
0:11:56	that feeding in position or some my
0:11:59	i is question so it can this work out that you've the solution
0:12:02	even in the presence of such random in practice
0:12:05	some of the no
0:12:07	we consider a particular to so to make
0:12:09	no
0:12:10	a round of a remote
0:12:12	so we model the graph
0:12:14	right in in rash
0:12:15	i'm the users as a random graph
0:12:18	so
0:12:18	considering that large
0:12:20	my
0:12:20	right
0:12:21	row yeah this sequence
0:12:23	of a laplacian matrix
0:12:24	where we uh can be decomposed the sum of the next i R
0:12:29	the right
0:12:30	row
0:12:30	and
0:12:31	is a sequence
0:12:32	zero we i
0:12:35	moreover
0:12:36	we consider that
0:12:37	each you are not communication channel uses a you
0:12:40	either
0:12:41	a into them is a zero means
0:12:42	and
0:12:43	uniform stupid
0:12:44	ever
0:12:45	you particular or however condition and and he is one errors
0:12:49	i
0:12:50	and the how how it happens
0:12:52	what's so for consensus on
0:12:53	it it can be once the convert
0:12:55	are
0:12:56	so what would you throw but
0:12:58	we consider it a proposition
0:12:59	which allows
0:13:00	they but decision error with some use
0:13:03	a statistical properties that we
0:13:05	for to program
0:13:08	so
0:13:09	substituting in the expression on uh of the but i'm version
0:13:13	forming out
0:13:14	we we can write a partial in this way where now
0:13:18	visions
0:13:19	that
0:13:20	right
0:13:21	things
0:13:22	the and pay
0:13:23	because they are read station
0:13:24	random bar
0:13:26	and the ne K and that's O K A are now that to contribution
0:13:30	a quantization noise
0:13:31	and of
0:13:32	uh
0:13:33	and one of the
0:13:36	so a now on the for
0:13:38	of the function G of try out function G
0:13:42	yeah a or of
0:13:44	they then now
0:13:45	of the overall system be expressed in compact for you know we and way
0:13:49	for this he
0:13:50	can be re than
0:13:51	and the written as the sum
0:13:53	oh but that there is a function of are that depends on the me graph
0:13:56	on the spectrum
0:13:58	and are are a function yeah
0:14:00	that the base actually on the
0:14:03	a random vectors
0:14:04	and on the contribution of quantisation noise
0:14:08	um
0:14:09	what so what they do of the day what you want to say is that is where we got it
0:14:13	can be seen people uh
0:14:15	are a bit more also think approximation procedure
0:14:17	and find its so if their basic function are
0:14:20	with but you is met at each time he's
0:14:23	and this brought by
0:14:25	random
0:14:26	yeah
0:14:26	so what proved
0:14:28	is that giving this competence
0:14:30	which
0:14:31	that
0:14:31	these are quite common
0:14:33	think
0:14:34	additionally
0:14:35	a a a sequence that satisfies
0:14:37	addition
0:14:38	for
0:14:39	in english
0:14:40	sides
0:14:41	that
0:14:42	we can assure also
0:14:43	number
0:14:44	oh
0:14:45	that's what we got
0:14:46	a a so it's probably one
0:14:48	basic well
0:14:49	all
0:14:50	they
0:14:51	a of this work out we with a very to this
0:14:54	so we set E
0:14:56	and a social solution set composed of
0:14:58	zero
0:14:59	of these that their function
0:15:01	ah
0:15:03	at the rest of course on me
0:15:07	now a a relation to show
0:15:11	mary
0:15:12	but it you power
0:15:13	a results
0:15:14	so we can see there's more me
0:15:16	frequency domain
0:15:17	present
0:15:18	five
0:15:19	so of course in a really communication scenarios i
0:15:22	maybe there
0:15:24	because of
0:15:24	but
0:15:25	not
0:15:26	so we consider a these where they runs but it's are problem
0:15:29	without requiring sort transmission
0:15:31	and source you find the right the face
0:15:33	all uh each no
0:15:36	in the figure we show that be if you're of the normalized
0:15:38	simple
0:15:39	but
0:15:40	with respect to the iteration in
0:15:42	four
0:15:42	if if to what to establish a
0:15:46	a mission
0:15:47	read read is the i case
0:15:48	green zero point than i zero point seven zero
0:15:51	fine
0:15:52	as we can see big vector
0:15:53	but that that the call
0:15:55	yeah
0:15:55	as will always reached
0:15:57	we a shown that the and don't expect
0:15:59	row
0:16:00	oh that are what we can see is that to reduce it will be a stuff should actually
0:16:05	the network requires a longer
0:16:07	to reach the final
0:16:10	so
0:16:11	we propose this this through with the resource allocation strategy
0:16:15	me
0:16:16	the
0:16:17	for two if
0:16:19	this form and
0:16:20	a we you an S you know what fraction of also forces
0:16:24	a
0:16:25	to the Q five
0:16:26	a was lost
0:16:27	use
0:16:28	what would be put i'm second the users
0:16:31	you particular each node is supposed to be able to and only to nearby ones
0:16:35	and this problems
0:16:35	intrinsic capability
0:16:37	to provides also used
0:16:39	channel
0:16:40	to a real
0:16:41	for
0:16:42	and
0:16:43	at the end we have seen that the method is also about a packet drops
0:16:47	with that that is only to slow down to convert
0:16:52	can
0:16:53	i
0:17:04	or
0:17:08	uh_huh
0:17:09	well well what we have a a a a a a kind of a simulation all
0:17:13	a we what we what we what we do
0:17:15	select
0:17:17	and
0:17:18	press
0:17:18	large now
0:17:19	okay
0:17:20	comp
0:17:21	to
0:17:22	and no so that the for a mean it will not and
0:17:25	with
0:17:26	channel
0:17:27	of course
0:17:27	we need
0:17:28	so
0:17:28	number of channel
0:17:29	oh each
0:17:31	that's
0:17:31	and there are no
0:17:33	sparse
0:17:39	oh

BIO-INSPIRED SWARMING MODELS FOR DECENTRALIZED RADIO ACCESS INCORPORATING RANDOM LINKS AND QUANTIZED COMMUNICATIONS

Bio-inspired Information Processing and Networks

Presented by: Paolo Di Lorenzo, Author(s): Paolo Di Lorenzo, Sergio Barbarossa, University of Rome La Sapienza, Italy