NetworkAdaptive Video Streaming over Wireless Multihop Networks: HopbyHop Video Rate Control

1
Network-Adaptive Video Streaming over Wireless
Multi-hop Networks Hop-by-Hop Video Rate Control
Lab Seminar

• 18th, January, 2008
• SangHoon Park
• shpark_at_nm.gist.ac.kr
• Networked Media Laboratory
• Dept. of Information Communications
• Gwangju Institute of Science Technology

2
Contents

• Introduction
• Motivation
• Hop-by-Hop Network-Adaptive Video Rate Control
• Experimental Results
• Conclusions

3
Introduction

• Wireless Multi-hop Networks (WMN)
• Cheap and efficient network connectivity in a
  large region
• Real-time Video Streaming Service over WMN
• In recent, significant attention in both research
  and industrial domains
• Representative Challenges to provide High-Quality
  Video Service
• Limited and time-varying network available
  bandwidth
• Intra-/-inter flow interference, hidden terminal
  problem,
• Large delay due to frequent disconnection and
  change of video path
• Higher bit error rate (BER)

video streams
Video receiver
Internet
Video Server
Video receiver
Wireless Multi-hop Networks
Video receiver
Video receiver
4
Research Category

• Efficient and resilient video coding and
  protection
• Various error-resilience tools localization,
  data partitioning, redundant coding, and
  concealment-driven
• Coding Schemes for multi-path streaming
• Feedback-based Reference Picture Selection (RPS)
• Layered Coding with Selective ARQ (LS-SA)
• Multiple Description Coding (MDC)
• Effective network/transport protocol for video
  delivery
• Multi-path streaming
• Advantages Higher aggregate bandwidth, Reduction
  of the short term correlation in real-time
  traffic, and Reduction the chance of interrupting
  
• Issues
• Packet losses due to different causes should be
  differentiated
• - Congestion loss, channel error, route
  change/break
• To choose multiple maximally disjointed paths and
  assign appropriate sending rate
• Proposed protocols
• ADTFRC, MRTP/MRTCP, RMPSR, AMTP,

5
Research Category (Cont.)

• Link-layer scheduling for video delivery
• Congestion-distortion optimized scheduling
  (CoDiO)
• To determine a scheduling minimizing the expected
  Lagrangian cost D??
• D distortion of the received video stream
• ? the end-to-end-delay
• Layer 2.5 MAC for video delivery
• QoS-enhanced 802.11 MAC (e.g., EDCF) does not
  provide adequate service differentiation in WMN
  due to the hidden terminal and other interference
  problems
• Layer 2.5 SoftMAC
• coarse-grained control to coordinate and
  regulate network load and packet transmission of
  both real-time and best-effort traffic among
  neighbor nodes in a distributed manner.
• Cross-layer design for multi-hop video delivery
• Resource management, adaptation, and protection
  strategies in the lower layers (PHY, MAC, and
  Network/Transport Layers) are optimized without
  explicitly considering the specific
  characteristic of the multimedia applications
• E.g., Routing Video Coding, Power control
  Scheduling

6
Problem Description addressed in this work

• For individual video stream
• To improve end-to-end video quality by
  controlling video flow (sending rate), given
  time-varying WMN condition (e.g.,
  intra-/inter-flow interferences)
• Relation with Congestion control Our concern is
  end-to-end video quality of individual video
  stream rater than network. However, Closely
  related to each other.
• Assumption
• Single-path video streaming
• Multi-path approach has still some drawbacks
  (e.g., MDC has high decoding complexity and low
  coding efficiency)

Background traffic sender
Background traffic sender
Video Streaming path 1
Video receiver
Video sender
Video Streaming path 2
Video receiver
Background traffic receiver
Background traffic sender
Background traffic receiver
7
Important Issues needs to be addressed

• Accurate Network Condition (or Congestion)
  Detection
• In wired networks, the cause of packet loss is
  mainly congestion
• In wireless multi-hop networks, however, many
  causes (e.g., congestion, link error, route
  break, )
• False Congestion Detection results in congestion
  collapse or video quality distortion
• Fast Network Condition Detection
• In severely congested network condition,
  delivery of network and wireless channel
  monitoring information can be severely delayed
• Consideration of video quality in rate control
• Video units (e.g., frames) have dependency
• Local Vs Global Quality Optimization
• For multiple-video streams (or users), global
  quality needs to be defined and should be
  optimized
• Fairness with best-effort traffics

8
Approaches Related Work

• Two approaches to detect network congestion
• End-to-end measurement
• e.g., TCP, TFRC
• Advantage end-to-end semantics of TCP,
  convenient implementation, no infrastructure
  support
• Feedback from intermediate nodes (or
  network-assisted approach)
• random early detection/explicit congestion
  notification (RED/ECN)
• Advantage more direct monitoring of congestion
• TCP Friendly Rate Control (TFRC)
• Useful for video streams and provides fairness
  with best-effort traffics
• End-to-end approach
• End-to-end measurement Round-trip-time (RTT),
  packet loss rate
• End-to-end feedback
• In wireless multi-hop networks
• Needs to be modified (e.g., Differentiation of
  packet loss causes)
• 03_Fu
• Multi-metric joint identification approach
  inter-delay difference (IDD) and short term
  throughput (STT)

9
Approaches Related Work (Cont.)
(a) Arrival time of end-to-end feedback packets

• Two approaches to control video sending rate
• Sender-driven control
• e.g., TCP, TFRC
• Hop-by-hop control 07_Yuang
• Typically, hop feedback is based on queue length
  in each node
• This approach argues that hop-by-hop control
  schemes react to congestion faster than
  end-to-end schemes (the bottleneck node would
  send feedback backward, thus decreasing the delay
  in the control loop)
• Drawbacks
• Deployment Internet congestion control has been
  dominated by end-to-end schemes (e.g., TCP)
• Require to have per-flow state management in
  intermediate nodes
• - They argue that the number of flows per node
  is small in WMN

(b) Playout discontinuity
Video Streaming path 1
Hop feedback
Hop feedback
Hop feedback
10
Proposed Approach

• Current status on hop-by-hop control
• Existing hop-by-hop control schemes have been
  focused on the congestion control
• Recently, few papers addressed video issues
• Proposed approach Hop-by-hop video rate control
  based on hop feedback
• Intermediate nodes monitors channel condition of
  local link and back-propagate it
• Intermediate nodes adapt video sending rate to
  the channel condition of bottleneck link by
  comparing the local channel condition and hop
  feedback information

Background traffic
n
11
Wireless Network Monitoring Module

• Loss rate of video stream at the local channel is
  measured
• The ratio of the number of discarded video
  packets at MAC layer interface queue over the
  number of total video packets arrived at the
  queue
• For this, a per-flow state table is maintained

12
Basic Algorithms Video Rate Adaptation and
Feedback Signaling Modules
Video adaptation module
13
Priority-based Packet Dropping Module

• For given kn,
• Discards video packets with low priority
• E.g., temporal scalability I gt P gt B frame
  packets
• Priority field needs to be inserted
• E.g., A video stream with MPEG-2 TS over RTP
  profile
• Frame indexing
• Priority (temporal layer information)

13
14
Experimental Setup
video streaming path
KOREN

• Deployed in GIST DIC 2nd floor
• 1 Gateway (N1), 6 Intermediate nodes (N2N7)
• IEEE 802.11a-based single interface

N1 (gateway)
N2
N3

• Experimental video
• GOP IBBPBB, 30fps, 4Mbps
• 4 Temporal layers (l1, l2, l3, l4)
• l1 1.52Mbps, l2 0.86Mbps, l3 0.8Mbps, l4
  0.8Mbps
• Frame rate profile of each temporal layer
• l1 5fps, l2 5fps, l3 10fps, l4 10fps

N4
N5
N6
video receiver
15
Experimental Results

• For end-to-end quality assessment, variance of
  discontinuity (sd )is measured

sd (sec2)
Experiment number
16
Conclusions

• We proposed a hop-by-hop network adaptive video
  streaming scheme in wireless multi-hop networks
• The proposed scheme is based on cross-layer
  design
• In future work, multiple video streaming and
  fairness issue with non-realtime traffics need to
  be addressed

17
Reference

• 03_Fu Z. Fu, et al., A transport protocol for
  supporting multimedia streaming in mobile ad hoc
  networks, IEEE JSAC 2003, pp. 1615-1626, Dec.
  2003.
• 07_Yung Y. Yi and S. Shakkottai, Hop-by-hop
  congestion control over a wireless multi-hop
  network IEEE/ACM TON, vol. 15, no. 1, pp.
  133-144, Feb. 2007.

18

• Thanks !
• QA