Author
dangkhanh
View
221
Download
0
Embed Size (px)
1
2014/10/28 Zhisheng Niu @ Tsinghua University 1
Collaborative and Opportunistic Scheduling
in Mobile Ad Hoc Networks
for Autonomous Multi-robotic Systems
Zhisheng NiuTsinghua National Laboratory for Information Science and Technology
Tsinghua University, Beijing100084
2014.10.28 @ University of Hamburg
2014/10/28 Zhisheng Niu @ Tsinghua University 2
Content
Autonomous Multi-robotic Systems
WLAN, WPAN, and Mobile Ad Hoc Networks
A Collaborative and Opportunistic Scheduling Scheme
for Mobile Ad Hoc Networks
Summary
2
Why Robots Communicate?
Robots make decisions based on their sensing without coordination, two robots could decide conflicting
actions based on their sensing
Robot Coordination using Communication if two robots see the ball, they could both decide to go to the
ball. Instead, by communicating what they sensed, they can coordinate their roles and actions
Other examples: 1) Pushing a large object with multiple robots; 2) Searching a room for an object
2014/10/28 Zhisheng Niu @ Tsinghua University 3
1. Both robots see the ball.
3
2. Each robot communicates to the other its distance to the ball as seen by its own vision.
3. The robot closest to the ball becomes the Attacker.
4. The other robot becomes the Defender.
4
5. The Attacker approaches the ball.
6. The Defender stays in position.
7. The Attacker gets ready to kick the ball.
8. The Defender heads to a defensive position.
5
How to Communicate?
Infrastructure-based (centralized, coordinated) Robots send sensed information to a coordinator or
controller
Coordinator/Controller fuses the information, makes decision, and assign actions to robots
Generally good performance, but not scalable and robust
Infrastructure-less (distributed, autonomous) A robot sends its sensed information and his/her role (and
also action?) that he/she wants to take to all their neighbors
The neighbors received the information, take their own roles and actions, and broadcast to their neighbors (?)
Autonomous and robust, but may not optimal in performance. Also, communication overhead may be large
2014/10/28 Zhisheng Niu @ Tsinghua University 9
1. One robot sees the ball.
6
2. The robot seeing the ball communicatesactions to the other and assigns roles.
3. The robot that sees the ball is the Attacker.
4. The other robot is the Defender.
7
5. The Attacker heads to the ball.
6. The Defender moves aside to clear the shot to the goal.
What’s the Problem with Communication?
What kind of communications are being used? Cellular?
WiFi?
Bluetooth?
ZigBee?
Ultra Wide Band?
What are the problems? Packet loss?
Out of connection?
Delay?
Battery life?
Any new requirements? Security?
Reliability?
2014/10/28 Zhisheng Niu @ Tsinghua University 14
8
15
Two Types of Wireless Networks
channelWireless channel
BS/AP
A B
Infrastructure-based Network
(WLAN, WiMAX)
C D
Multi-hop Network(Ad hoc networks, mesh
networks, sensor networks)
Infrastructure-based Cellular Networks
2014.02.24 牛志升@清华大学 16
9
Infrastructure-based Cellular Networks
2014.02.24 17牛志升@清华大学
2020 5G?(BDMA? Massive MIMO) All-IP, U-shape traffic 100-1000Mb/s
(FDMA)
(TDMA)
(CDMA)
(OFDMA, MIMO)
01011011
InternetAccessPoint
0101 1011
• 802.11b– Standard for 2.4GHz ISM band (80 MHz)– Direct sequence (DS) spread spectrum
• 802.11a– Standard for 5GHz NII band (300 MHz)– OFDM with time division– Similar to HiperLAN in Europe
• 802.11g– Standard in 2.4 GHz band with 54Mbps– OFDM
• 802.11n– Standard in 2.4 GHz or 5GHz band– OFDM + MIMO– Speeds up to 250 Mb/s (110 Mb/s
in practice) and about twice the range
Infrastructure-based WLAN
2014.02.24 牛志升@清华大学 18
10
Infrastructure-less WPAN
Bluetooth (initiated by Ericsson, Danish King’s nick name, IEEE 802.15.1)
Short range connection (10-100 m) in 2.4GNz ISM band
1 data (721 Kbps) and 3 voice (56 Kbps) channels
Bluetooth 2.0/3.0 for higher data rate, 4.0 for low-power
ZigBee (initiated by Honeywell, Bee’s Waggle Dance, IEEE802.15.4)
Low-Rate WPAN (20, 40, 250 Kbps), but very low power consumption and low cost over ISM bands
Mainly used for large-scale sensor networks
Ultra WideBand (UWB) (initiated by Motorola, Pulse Communication)
Impulse radio (no carrier): sends pulses of tens of picoseconds to nanoseconds (Low probability of detection high reliability/security)
Uses a lot of bandwidth (GHz) to achieve very high data rates (~100 Mbps), low delay, and low power
Low range, 10m or less, due to power restriction
Mainly used for high-speed comm and localization2014.02.24 19牛志升@清华大学
Xbee (ZigBee-based but extended, Licensed by DiGi)
Platform XBee XBee-PROPerformance
Power output1mW (+0 dBm) North American & International version
63 mW (+18 dBm) North American version 10 mW (+10 dBm) International version
Indoor/Urban range Up to 100 ft (30 m) Up to 300 ft (90 m)Outdoor/RF line-of-sight range Up to 300 ft (90 m) Up to 1 mile (1.6 km) RF LOSReceiver sensitivity -92 dBm -100 dBm (all variants)RF data rate 250 Kbps 250 KbpsOperating frequency 2.4 GHz 2.4 GHzInterface data rate Up to 115.2 Kbps Up to 115.2 Kbps
NetworkingSpread spectrum type DSSS (Direct Sequence Spread Spectrum)Supported network topologies Point-to-point, point-to-multipoint, & peer-to-peerError handling Retries & acknowledgementsFiltration options PAN ID, Channel, and 64-bit addressesChannel capacity 16 Channels 12 ChannelsAddressing 65,000 network addresses available for each channel
Power
Supply voltage2.8 - 3.4 VDCXBee Footprint Recommendation: 3.0 - 3.4 VDC
2.8 - 3.4 VDCXBee Footprint Recommendation: 3.0 - 3.4 VDC
Transmit current45 mA (@ 3.3 V) boost mode 35 mA (@ 3.3 V) normal mode
215 mA (@ 3.3 V)
Receive current 50 mA (@ 3.3 V) 55 mA (@ 3.3 V)Power-down sleep current <10 µA at 25° C <10 µA at 25° C
GeneralFrequency band 2.4000 - 2.4835 GHz
Physical PropertiesSize 0.960 in x 1.087 in (2.438 cm x 2.761 cm) 0.960 in x 1.297 in (2.438 cm x 3.294 cm)Weight 0.10 oz (3g)Antenna options U.FL, Reverse Polarity SMA (RPSMA), chip antenna or wired whip antennaOperating temperature -40° C to 85° C (industrial) 20
11
Infrastructure-less Mobile Ad-Hoc Network
Peer-to-peer communications. No backbone infrastructure. Routing can be multihop and multipath
Network topology is dynamic (Extended concept of mobility: network mobility or moving routers)
2014.02.24 21牛志升@清华大学
Mobile Ad Hoc Networks
Self-organized multihop multipath communications
12
Wireless Mesh Networks
Multi-layer Wireless Communication for Large-Scale Mobile Ad Hoc Networks
2014/10/28 Zhisheng Niu @ Tsinghua University 23
MESH
Wireless Sensor Networks
Nodes powered by non-rechargeable batteries
Data flows to centralized location (sink)
Low per-node rates but up to 100,000 nodes.
Data highly correlated in time and space.
Nodes can cooperate in transmission, reception, compression, and signal processing.
2014.02.24 24牛志升@清华大学
cluster cluster cluster
sinkHead
Head Head
13
2007-06-21 Zhisheng Niu @ Tsinghua University 25
Distributed Coordination Function (Virtual carrier sensing)
Sender sends Ready-To-Send (RTS)
Receiver sends Clear-To-Send (CTS)
RTS and CTS reserves the area around sender and receiver for the duration of dialogue
Nodes that overhear RTS and CTS defer transmissions by setting Network Allocation Vector (NAV)
MAC Protocol for Contention/Collision Avoidance
2007-06-21 Zhisheng Niu @ Tsinghua University 26
802.11 Distributed Coordination Function
A
B
C
D
A B C D
Time
14
2007-06-21 Zhisheng Niu @ Tsinghua University 27
802.11 Distributed Coordination Function
A
B
C
D
A B C D
RTS
Time
RTS
2007-06-21 Zhisheng Niu @ Tsinghua University 28
802.11 Distributed Coordination Function
A
B
C
D
A B C D
RTS
CTS
SIFS
NAV Time
CTS
15
2007-06-21 Zhisheng Niu @ Tsinghua University 29
802.11 Distributed Coordination Function
A
B
C
D
A B C D
RTS
CTS
DATA
SIFS
NAV
NAV
Time
DATA
SIFS: Short Inter-Frame Space
2007-06-21 Zhisheng Niu @ Tsinghua University 30
802.11 Distributed Coordination Function
A
B
C
D
A B C D
RTS
CTS
DATA
SIFS
ACK
NAV
NAV
Time
ACK
16
2007-06-21 Zhisheng Niu @ Tsinghua University 31
802.11 Distributed Coordination Function
A
B
C
D
A B C D
RTS
CTS
DATA
SIFS
ACK
NAV
NAV
DIFS
Time
Contention Window
DIFS: DCF Inter-Frame Space
2007-06-21 Zhisheng Niu @ Tsinghua University 32
802.11 Power Saving Mechanism
Time is divided into beacon intervals
All nodes wake up at the beginning of a beacon interval for a fixed duration of time (ATIM window)
Exchange ATIM (Ad-hoc Traffic Indication Message) during ATIM window
Nodes negotiate channels using ATIM messages Nodes that received ATIM message stay up during for the whole beacon
interval
Nodes that do not receive ATIM message may go into doze mode after ATIM window
Nodes switch to selected channels after ATIM window for the rest of the beacon interval
17
2007-06-21 Zhisheng Niu @ Tsinghua University 33
802.11 Power Saving Mechanism
A
B
C
Time
Beacon
ATIM Window
Beacon Interval
2007-06-21 Zhisheng Niu @ Tsinghua University 34
802.11 Power Saving Mechanism
A
B
C
Time
Beacon
ATIM
ATIM Window
Beacon Interval
18
2007-06-21 Zhisheng Niu @ Tsinghua University 35
802.11 Power Saving Mechanism
A
B
C
Time
Beacon
ATIM
ATIM-ACK
ATIM Window
Beacon Interval
2007-06-21 Zhisheng Niu @ Tsinghua University 36
802.11 Power Saving Mechanism
A
B
C
Time
Beacon
ATIM
ATIM-ACK
ATIM-RES
ATIM Window
Beacon Interval
19
2007-06-21 Zhisheng Niu @ Tsinghua University 37
802.11 Power Saving Mechanism
A
B
C
Time
Beacon
ATIM
ATIM-ACK
DATAATIM-RES
Doze Mode
ATIM Window
Beacon Interval
2007-06-21 Zhisheng Niu @ Tsinghua University 38
802.11 Power Saving Mechanism
A
B
C
Time
Beacon
ATIM
ATIM-ACK
DATA
ACK
ATIM-RES
Doze Mode
ATIM Window
Beacon Interval
20
2007-06-21 Zhisheng Niu @ Tsinghua University 39
Channel Negotiation (multi-channel case)
A
B
C
DTime
ATIM Window
Beacon Interval
Common Channel Selected Channel
Beacon
2007-06-21 Zhisheng Niu @ Tsinghua University 40
Channel Negotiation
A
B
C
D
ATIM
ATIM-ACK(1)
ATIM-RES(1)
Time
ATIM Window
Beacon Interval
Common Channel Selected Channel
Beacon
21
2007-06-21 Zhisheng Niu @ Tsinghua University 41
Channel Negotiation
A
B
C
D
ATIM
ATIM-ACK(1)
ATIM-RES(1)
ATIM-ACK(2)
ATIM ATIM-RES(2)
Time
ATIM Window
Beacon Interval
Common Channel Selected Channel
Beacon
2007-06-21 Zhisheng Niu @ Tsinghua University 42
Channel Negotiation
A
B
C
D
ATIM
ATIM-ACK(1)
ATIM-RES(1)
ATIM-ACK(2)
ATIM ATIM-RES(2)
Time
ATIM Window
Beacon Interval
Common Channel Selected Channel
Beacon
RTS
CTS
RTS
CTS
DATA
ACK
ACK
DATA
Channel 1
Channel 1
Channel 2
Channel 2
22
2007-06-21 Zhisheng Niu @ Tsinghua University 43
RTS/CTS solves Hidden/Expose Terminal Problems
A B CDATA
C does not hear A’s transmission
Hidden Terminal Problem
2007-06-21 Zhisheng Niu @ Tsinghua University 44
Hidden Terminal Problem
A B CDATA
C starts transmitting – collides at B
23
2007-06-21 Zhisheng Niu @ Tsinghua University 45
Solution: Virtual Carrier Sensing
A B CRTS
A sends RTS
D
D overhears RTS and defers transmission
2007-06-21 Zhisheng Niu @ Tsinghua University 46
Solution: Virtual Carrier Sensing
A B CCTS
B sends CTS
D
C overhears CTS and defers transmission
24
2007-06-21 Zhisheng Niu @ Tsinghua University 47
Solution: Virtual Carrier Sensing
A B CDATA
D
A sends DATA to B
2007-06-21 Zhisheng Niu @ Tsinghua University 48
Solution: Virtual Carrier Sensing
A B CRTS
D
D overhears RTS and defers transmission
25
Route Discovery and Establishment
• Automatic route discovery and establishmentA neighbor either broadcasts the RREQ to its neighbors or satisfies the RREQ by sending a RREP back to the source
Later copies of the same RREQ request are discarded
Node records the address of the sender of RREQ
Entries are discarded after a time-out period
Eventually, a RREQ arrives at a node that possesses the current route for the destination (Comparison of sequence numbers)
Node unicasts a route reply packet (RREP) back to the neighbor from which it received the RREQ. The RREP travels along the path established in the reverse path set-up
Each node along the RREP journey sets up a forward pointer, updates time-out entries, records the destination sequence number of requested destination
2014/10/29 49牛志升@清华大学
Example: AODV (Ad-hoc On-demand Distance Vector)
Routing Scheme
B
S
E
C G
F
A
H
D
Y
I
K
P
L
J
TZ
RREQ
2014/10/29 50牛志升@清华大学
26
AODV: Example
B
S
E
C G
F
A
H
D
Y
I
K
P
L
J
TZ
Reverse Path
Setup2014/10/29 51牛志升@清华大学
AODV: Example
B
S
E
C G
F
A
H
D
Y
I
K
P
L
J
TZ
2014/10/29 52牛志升@清华大学
27
AODV: Example
B
S
E
C G
F
A
H
D
Y
I
K
P
L
J
TZ
2014/10/29 53牛志升@清华大学
AODV: Example
B
S
E
C G
F
A
H
D
Y
I
K
P
L
J
TZ
RREP
2014/10/29 54牛志升@清华大学
28
AODV: Example
B
S
E
C G
F
A
H
D
Y
I
K
P
L
J
TZ
Forward Path
Setup
2014/10/29 55牛志升@清华大学
AODV (Example)
B
S
E
C G
F
A
H
D
Y
I
K
P
L
J
TZ
2014/10/29 56牛志升@清华大学
29
AODV (Example)
B
S
E
C G
F
A
H
D
Y
I
K
P
L
J
TZ
2014/10/29 57牛志升@清华大学
AODV (Example)
B
S
E
C G
F
A
H
D
Y
I
K
P
L
J
TZ
2014/10/29 58牛志升@清华大学
30
59
Routing Algorithms in MRMC Networks
In MRMC, the shortest path algorithm does not work well Different radios may have different characters Interference among radios is not considered in shortest path
Need to exploit channel diversity Selecting channel diverse routes
Interface switching cost has to be considered Switching interfaces incurs a delay A node may be on different routes, requiring switching
A
B
C
D
1 1
2 1
Route A-C-D is betterWhen possible, select routes where
different hops are on different channelsRoute A-B-D is better
A
B
C
D2 1
2 1
E3
When possible, select routes that do not require frequent switching
2014/10/28 Zhisheng Niu @ Tsinghua University
Opportunistic Scheduling
Multi-User Diversity In a multiuser system, channels
(timeslots/subcarriers/antennas) that are poor to one user may be the best for others
Multi-Link Diversity Scheduling the link that has the
instantaneously best channel condition would increase the system and link capacity.
Opportunistic Scheduling
f
f
f
f
f
f
2014/10/28 60Zhisheng Niu @ Tsinghua University
31
Dynamics in Wireless Channels and Diversities
Wireless channel varies dramatically over time/frequency/space Interferences, Fading, Multipath, Doppler, …
Traditional view treats the channel uncertainty as obstacle and therefore try to avoid it e.g., interleave, AMC (adaptive modulation and coding)
Modern view treats the channel variation as opportunity and therefore try to exploit it Diversity: always choose the best frequency/time/antenna to transmit and leave
out the poor ones e.g., OFDM, MIMO
channel
2014/10/28 61Zhisheng Niu @ Tsinghua University
Key Challenges for Exploiting MU Diversity in MANET
Absence of a central scheduler: Transmitter Cooperation!
Neighboring transmitters should jointly determine the “on-peak” flows
Transmitter should offer more transmission opportunities to the flow which has not achieved its QoS requirement
Neighboring transmitters should be coordinated to reserve the shared wireless bandwidth to reduce the potential collision to that flow
Absence of a dedicated channel for feedback A complement scheme needs to be designed, which has to avoid collisions
and to have an acceptable overhead
2014/10/28 Zhisheng Niu @ Tsinghua University 62
32
The Existing Opportunistic Algorithms
2014/10/28 Zhisheng Niu @ Tsinghua University 63
Exploit Time Diversity only but not Multi-user Diversity
Auto Rate Fallback (ARF)
ARF protocol attempt to use higher transmission rates after consecutive transmission successes, and revert to lower rates after failures.
Receiver Based Auto Rate (RBAR)
RBAR using physical-layer analysis of the RTS message determines the rate by receiver.
Opportunistic Auto Rate (OAR) protocol
OAR [1] opportunistically send multiple back-to-back data packets whenever the channel quality is good.
[1] B. Sadeghi, V. Kanodia, A. Sabharwal, and E. Knightly. “Opportunistic Media Access for Multirate Ad Hoc Networks,” Proc. ACM MOBICOM’ 02,2002.
The Existing Opportunistic Algorithms
2014/10/28 Zhisheng Niu @ Tsinghua University 64
[3] J. Wang, H. Zhai, Y. Fang, J. M. Shea and D. Wu, "OSAR: Utilizing Multiuser Diversity in Wireless Ad Hoc Networks," IEEE Trans. on Mobile Computing, vol. 5, no. 12, pp. 1764-1779, Dec. 2006.
MAD: Media Access Diversity[2]
OSAR: Opportunistic packet-scheduling and auto-rate[3]
A sender broadcasts a probing multicast/group RTS packet, then its receivers measure the SINR and send back the result by CTS packets. The sender then selects the receiver with the best channel condition.
[2] Z. Ji, Y. Yang, J. Zhou, M. Takai, R. Bagrodia, “Exploiting medium access diversity in rate adaptive wireless LANs,”, in Proc. ACM MobiCom’04, pp. 345–359, Sept. 2004.
Exploiting multi-user diversity but not multi-link diversityNo cooperation among neighboring transmitters (local optimization only without QoS guarantee)
33
If without cooperation,
If A chooses B, and then F chooses G by local scheduling, link A-B is corrupted by hidden terminal F in link F-G.
If a bandwidth requirement is assigned to the flow A-B, this link suffers a high collision probability induced by hidden terminal F.
Why Cooperative Scheduling?
2014/10/28 65Zhisheng Niu @ Tsinghua University
Cooperative & Opportunistic Scheduling with QoS Constraints
Exploit time-diversity and multiuser diversity simultaneously, while providing QoS guarantee
Through cooperation, some transmissions are deferred to favor some other links which have not achieved their QoS requirements
2014/10/28
Our Solution: QoS-Aware COS
Zhisheng Niu @ Tsinghua University 6666
TIFS: Traffic-control Inter-Frame Space[1] Q. Zhang, Q. Chen, F. Yang, X. Shen, Z. Niu, "Cooperative and Opportunistic Transmission for IEEE 802.11-based Ad Hoc Networks,"IEEE Networks, vol.21, No.1, pp.14-20, 2007.[2] Q. Chen, Q. Zhang, Z. Niu, "QoS-aware Cooperative and Opportunistic Scheduling Exploiting Multi-user Diversity for Rate AdaptiveAd Hoc Networks," IEEE Trans. Vech. Tech., vol.57, no.2, pp.1113-1125, April 2008.[3] Q. Chen, Q. Zhang, Z. Niu, “A Graph Theory based Opportunistic Link Scheduling for Wireless Ad Hoc Networks”, IEEE Trans.Wireless Comm., Oct. 2009.
34
Problem Formulation
Cooprative & Opportunistic Scheduling (COS) with QoS requirements
Selection policy Q(t)Data rate of ith link
Indicator Function IXContention Function c(i, j, t)
Bandwidth Requirement Gi
…(1)
2014/10/28 67Zhisheng Niu @ Tsinghua University
Optimal Solution
Proposition 1: The optimal solution of opportunistic scheduling (1), if one exists, is of the following form.
in which Sm is an MIS, and λi’s are the KKT values.
Maximal Independent Set (MIS):= { S1, S2, S3, S4 } = { (F2), (F3), (F1,F4), (F1,F5) }
Conflict Graph
2014/10/28 68Zhisheng Niu @ Tsinghua University
(i.e., the link set selected by the optimal scheduling should be a MIS)
35
Here we focus on the minimum bandwidth constraints and the network throughput maximization,
in which
Therefore, the optimal criteria is: Credits
The KKT values can be calculated by stochastic approximation algorithm
Optimal Scheduling
2014/10/28 69Zhisheng Niu @ Tsinghua University
An Example
Considering the four MIS Ω={{F1,F5},{F2},{F3},{F4}}
Suppose that the credits of the four MIS are 7, 6, 5 and 4
Then, the credits of flows F1 to F5 are {7, 6, 5, 4, 7} and the credits of the transmitter A and D are both 7. flow's credit is set to the largest credit of the MIS's those include this
flow
transmitter's credit is set to the largest credit of the flows originated by this transmitter.
Then a set of flows, in one MIS with the largest credit, in this example flow 1 and 5, are scheduled to transmit simultaneously.
2014/10/28 Zhisheng Niu @ Tsinghua University 70
36
The Challenges of the Optimal Scheduling
The challenges of implementing the optimal scheduling in 802.11 based MANET
Exchanging parameters all over the network is impractical
Use 2-hop information exchanging & average data rates
Difficult to track the time-varying contention graph which is needed in the optimal scheduling
Use average Local Contention Graph (LCG)
To schedule a set of links in an ad hoc network in a deterministic pattern is not trivial
Insert an extra interval (TIFS) into consecutive data transmissions
2014/10/28 71Zhisheng Niu @ Tsinghua University
An Heuristic Adaptive TIFS Setting
Adaptive TIFS Setting Algorithm The optimal TIFS depends on
1) move pattern of the nodes2) number of the neighboring transmitters,3) contention graph,4) QoS requirements of each links.
An adaptive scheme:
2014/10/28 72Zhisheng Niu @ Tsinghua University
37
Maximal Weighted Independent Set (MWIS) problem
: an undirected graph: the weight of the vertex v: the sum of the weights of the vertices in X: the set of vertices adjacent to vertex v
MWIS: NP Hard!
vertex vedge
Vertex Links
Edges Contention Relationship
Weight or
A Distributive Approach to the Optimal Solution
2014/10/28 73Zhisheng Niu @ Tsinghua University
12
3
4
Weight
: 5
3+21 = 5
: 2.33
: 1.25
: 4
: Degree
Heuristic Algorithm:(1) select a minimal weighted degree vertex as a vertex in the weighted
independent set Sm(2) delete the vertex and all of its neighbors from the graph(3) repeat this process for the remaining subgraph until the graph becomes
empty.
Weighted Degree (WD):
Exchange weight and WD among neighboring links
A Distributed Heuristic Scheduling
2014/10/28 74Zhisheng Niu @ Tsinghua University
38
1. Probing the channel during RTS-CTS exchange2. Update the degree table by
1) probing process2) overhearing the CTS and DATA packets
3. Set the TIFS accordingly
Heuristic Scheduling
(Priority based Transmission)
2014/10/28 75Zhisheng Niu @ Tsinghua University
Group RTS
Heuristic Scheduling: An Example
Cooperative & Opportunistic Scheduling (COS)
2014/10/28 76Zhisheng Niu @ Tsinghua University
39
CTS replied &Parameter Exchanging(channel conditions & KKT multipliers)
D replies CTS.
E replies CTS.
C replies CTS.
Packet Scheduling
DATA transmission
ACK repliedNode A reset its TIFS.
Finally, we get two links those can transmit simultaneously!
Heuristic Scheduling: An Example
Cooprative & Opportunistic Scheduling (COS)
2014/10/28 77Zhisheng Niu @ Tsinghua University
Key point 1: How is the overhead?
LN (4bits): No. of Links (Entities) LN<=10LI (8bits): Link IdentifierDR (4bits): Data rate supported, 16 levelsDE (12bits): Weighted Degree of a link
Degree table
Heuristic Scheduling
2014/10/28 78Zhisheng Niu @ Tsinghua University
40
Heuristic Scheduling
Key point 2: how to set the TIFS?The optimal length of TIFS =
The expect time that the a sender will be scheduled from now.
Single Link Markov Chain:
Multi-Link Markov Chain:
N-Dimension Markov Process,N = the number of links.
2014/10/28 79Zhisheng Niu @ Tsinghua University
Heuristic Scheduling
Key point 2: how to set the TIFS?The optimal length of TIFS =
The expect time that the a sender will be scheduled from now.
Optimal TIFS:
2014/10/28 80Zhisheng Niu @ Tsinghua University
41
Heuristic Scheduling
Key point 2: how to set the TIFS?An example of the optimal length of TIFS of the first link
T0=10ms, v = 1m/s
2014/10/28 81Zhisheng Niu @ Tsinghua University
Heuristic Scheduling
Key point 3: what’s the performance bound?
While the average performance of the rate adaptive 802.11 MAC is
2014/10/28 82Zhisheng Niu @ Tsinghua University
42
2014/10/28 Zhisheng Niu @ Tsinghua University 83
No QoS req. G2=G3=1.5Mbps G2=G3=2.0Mbps
COS: Cooperative & Opportunistic Scheduling
OSAR: opportunistic w/o cooperation
OAR: standard multi-rate 802.11b
Simulation Results
2014/10/28
Grid Topology with hard QoS
Simulation Results
Zhisheng Niu @ Tsinghua University
COS: Cooperative & Opportunistic Scheduling
OSAR: opportunistic w/o cooperation
OAR: standard multi-rate 802.11b
84
43
2014/10/28 Zhisheng Niu @ Tsinghua University 85
Random Scenarios
Simulation Results
COS: Cooperative & Opportunistic Scheduling
OSAR: opportunistic w/o cooperation
OAR: standard multi-rate 802.11b
Conclusions
Proposed an interference-dependent multiuser diversity
model for MANET while considering the QoS requirements
of each flow
Provided an optimal criterion to find the global optimal set
of simultaneously transmitting flows, together with
a heuristic scheduling algorithm
Designed an IEEE 802.11-based QoS-aware distributed
cooperative and opportunistic scheduling (COS) scheme,
which obtains higher network throughput and better QoS
support than the existing schemes with limited local
information.
2014/10/28 86Zhisheng Niu @ Tsinghua University
44
2014/10/28 Zhisheng Niu @ Tsinghua University 87
Prof. Zhisheng Niu
Room 3-327, FIT Building
School of Information Science and Technology
Tsinghua University, Beijing 10084, China
Email: [email protected]
Tel: 86-10-62781423; Fax: 86-10-62773634
Thank You and Future Contact