Cross-Layer Design of Wireless Networks Andrea Goldsmith Stanford University wsl.stanford.edu Clean Slate Seminar Dec. 5, 2005

Future Network Applications Internet (for the Z generation) “Cellular” Entertainment Commerce Smart Homes/Spaces/Structures Sensor Networks Automated Highways/Factories … Applications have hard delay constraints, rate requirements, energy constraints, and/or security constraints that must be met These requirements are collectively called QoS

Challenges to meeting QoS Underlying channels, networks, and end-devices are heterogenous Traffic patterns, user locations, and network conditions are constantly changing Hard constraints cannot be guaranteed, and average constraints can be poor metrics. No single layer in the protocol stack can support QoS: cross-layer design needed

A Brief Introduction to Protocol Layers Premise: Break network tasks into logically distinct entities, each built on top of the service provided by the lower layer entities. Application Presentation Session Transport Network Datalink Physical Physical medium Example: OSI Reference Model

OSI vs. TCP/IP OSI: conceptually define services, interfaces, protocols Internet: provides a successful implementation Application Application Telnet FTP DNS Presentation Session TCP UDP Transport Transport IP Network Internet Datalink Host-to- network Packet radio LAN Physical OSI TCP/IP

Layer Functionality Application Transport Network Access Link Compression, error concealment, packetization, scheduling, … Transport End-to-end error recovery, retransmissions, flow control, … Network Neighbor discovery and routing Access Channel sharing, error recovery/retransmission, packetization, … Link Bit transmission (modulation, coding, …)

Layering Pros and Cons Advantages Disadvantages Simplification - Breaking the complex task of end-to-end networking into disjoint parts simplifies design Modularity – Protocols easier to optimize, manage, and maintain. More insight into layer operation. Abstract functionality –Lower layers can be changed without affecting the upper layers Reuse – Upper layers can reuse the functionality provided by lower layers Disadvantages Suboptimal: Layering introduces inefficiencies and/or redundancy (same function performed at multiple layers) Information hiding: information about operation at one layer cannot be used by higher or lower layers Performance: Layering can lead to poor performance, especially for applications with hard QoS constraints

Key layering questions How should the complex task of end-to-end networking be decomposed into layers What functions should be placed at each level? Can a function be placed at multiple levels? What should the layer interfaces be? Should networks be decomposed into layers? Design of each protocol layer entails tradeoffs, which should be optimized relative to other protocol layers What is the alternative to layered design? Cross-layer design No-layer design

Crosslayer Design: Information Exchange Across Layers Application Transport Network Access Link End-to-End Metrics Interdisciplinary research, design, and development very challenging, but necessary to meet the requirements of future wireless applications Substantial gains in throughput, efficiency, and QoS can be achieved with cross-layer design

Information Exchange Applications have information about the data characteristics and requirements Lower layers have information about network/channel conditions

Crosslayer Techniques Adaptive techniques Link, MAC, network, and application adaptation Resource management and allocation Diversity techniques Link diversity (antennas, channels, etc.) Access diversity Route diversity Application diversity Content location/server diversity Scheduling Application scheduling/data prioritization Resource reservation Access scheduling

Key Questions What is the right framework for crosslayer design? What are the key crosslayer design synergies? How to manage crosslayer complexity? What information should be exchanged across layers, and how should this information be used? How to balance the needs of all users/applications?

Cross-Layer Design Examples Joint source/channel coding in MIMO channels and networks Energy-constrained networks Distributed control over wireless networks

Joint S/C Coding with MIMO Use antennas for multiplexing Use antennas for diversity High-Rate Quantizer ST Code High Rate Decoder Error Prone Low Pe Low-Rate Quantizer ST Code High Diversity Decoder

Diversity/Multiplexing Tradeoff at High SNR‡ Define a family of block codes {C(SNR)} of length T with rate R(SNR)~r log SNR Given {C(SNR)}, define diversity and multiplexing gains asymptotically ‡Zheng/Tse 2002

Diversity/Multiplexing Tradeoffs Insert picture • Where do we operate on this curve? • Depends on higher-layer metrics

Bound on Total Distortion We use the high SNR bound (finite blocks) Equating source and channel rates and assuming high SNR (order terms negligible) As SNR, equate exponents to minimize distortion

Asymptotic Distortion With matching exponents, The distortion is a simple function of SNR and the optimal diversity/multiplexing point Asymptotically, leads to a familiar form Same form as Zheng/Tse Pe and rate formulas

Finite SNR Convex problem Solve using standard techniques Solution shows how to best use antennas

Distortion vs. Multiplexing N=M=8 k>>T

Distortion vs. Multiplexing N=M=8 k=O(T)

Joint Source/Channel Coding for Video The same optimization framework applies to a broad set of source/channel codes Progressive video encoding [Girod 2000], yields separable distortion DT=DS+DC Linear space-time codes [Kuhn 2003] that can tradeoff diversity and multiplexing Optimize # of antennas NU for multiplexing Optimization is an integer program

Antenna Assignment vs. SNR Can extend framework to include delay

Extensions Multiplexing/Diversity/Delay Tradeoffs Can also optimize throughput Throughput approaches capacity at high SNR Discontinuity at d=0 Tradeoffs under retransmissions Retransmissions add a third dimension of delay Region obtained by Caire, Damen, ElGamal (IT’05) Tradeoff optimization requires an end-to-end distortion measure that incorporates delay, rate, and Pe. Multiplexing versus diversity tradeoffs in networks Multiplexing/Diversity/Delay Tradeoffs

Delay/Throughput/Robustness across Multiple Layers Multiple routes through the network can be used for multiplexing or reduced delay/loss Application can use single-description or multiple description codes Can optimize optimal operating point for these tradeoffs to minimize distortion

Cross-layer protocol design for real-time media Loss-resilient source coding and packetization Application layer Rate-distortion preamble Congestion-distortion optimized scheduling Transport layer Congestion-distortion optimized routing Traffic flows Network layer Capacity assignment for multiple service classes Link capacities MAC layer Link state information Adaptive link layer techniques Link layer

Video streaming performance 5 dB 3-fold increase 100 1000 (logarithmic scale)

Ad-Hoc Networks Peer-to-peer communications. Fully connected with different link SINRs No centralized control Routing can be multihop. Topology is dynamic. -No backbone: nodes must self-configure into a network. -In principle all nodes can communicate with all other nodes, but multihop routing can reduce the interference associated with direct transmission. -Topology dynamic since nodes move around and link characteristics change. -Applications: appliances and entertainment units in the home, community networks that bypass the Internet. Military networks for robust flexible easily-deployed network (every soldier is a node).

Energy-Constrained Nodes Each node can only send a finite number of bits. TX energy minimized by sending each bit very slowly. Introduces a delay versus energy tradeoff for each bit. Short-range networks must consider both transmit and processing/circuit energy. Sophisticated techniques not necessarily energy-efficient. Sleep modes can save energy but complicate networking. Changes everything about the network design: Bit allocation must be optimized across all protocols. Delay vs. throughput vs. node/network lifetime tradeoffs. Optimization of node cooperation. All the sophisticated high-performance communication techniques developed since WW2 may need to be thrown out the window. By cooperating, nodes can save energy

Cross-Layer Optimization Model Min s.t. The cost function f0(.) is energy consumption. The design variables (x1,x2,…) are parameters that affect energy consumption, e.g. transmission time. fi(x1,x2,…)0 and gj(x1,x2,…)=0 are system constraints, such as a delay or rate constraints. If not convex, relaxation methods can be used. We focus on time division systems Joint work with S. Cui

Modulation Optimization

Minimum Energy Routing Transmission and Circuit Energy Red: hub node Blue: relay only Green: source 0.3 4 3 2 1 (0,0) (5,0) (10,0) (15,0) Multihop routing may not be optimal when circuit energy consumption is considered

Relay Nodes with Data to Send Transmission energy only 0.1 Red: hub node Green: relay/source 0.085 4 2 0.185 3 0.115 1 (0,0) (5,0) (10,0) (15,0) 0.515 • Optimal routing uses single and multiple hops • Link adaptation yields additional 70% energy savings

Cooperative MIMO Nodes close together can cooperatively transmit Form a multiple-antenna transmitter Nodes close together can cooperatively receive Form a multiple-antenna receiver Node cooperation can increase capacity, save energy, and reduce delay.

Capacity Gain vs Network Topology TX1 x1 x2 d=1 d=r<1 Cooperative DPC best Cooperative DPC worst RX2 Joint work with C. Ng

Cross-Layer Design with Cooperation Multihop Routing among Clusters

Total Energy versus Delay (with rate adaptation)

Cooperative Compression Source data correlated in space and time Nodes should cooperate in compression as well as communication and routing Joint source/channel/network coding

Energy-efficient estimation Sensor 1 g1 s22 g2 Sensor 2 Fusion Center gK s2K Sensor K Different channel gain (known) Different observation quality (known) Minimize the Pk’s for a given distortion constraint

Digital v.s. Analog Why doesn’t power associated with the fundamental limit in digital decrease with D_0?

Key Message Wireless ad hoc networks impose tradeoffs between rate, power/energy, and loss/delay The tradeoff implications for distributed control is poorly understood

Distributed Control over Wireless Links Joint work With X. Liu Packet loss and/or delays impacts controller performance There is little methodology to characterize this impact Controller sampling determines data rate requirements Network design and resulting tradeoffs of rate vs. loss and delay should be optimized for the controller performance

Optimal Controller under Packet Losses Shared Channel l1 l2 l3 x1 x2 Disturbance System Lossy Channel Lossy Channel Lossy Channel Control Command Optimal State Estimator State Feedback Controller State Estimate • This structure is optimal for LQG control with no losses. • Under lossy observations, prove that the optimal controller is a modified Kalman filter and state feedback controller. • The controller adapts to packet delay and loss, and its error covariance is stochastic • System stability depends on l1, l2, and l3 • These throughput parameters depend on the network design. One of our main contributions is considering control packet losses. Also, in time division the transmissions don’t interfere. Centralized control. Assume controller knows if a control command is received and the corresponding execution time. Controller knows the delay distribution of control commands and sensor data sent over the network With these assumptions, we prove that separation of the controller and state estimator is optimal.

Cross-Layer Design of Distributed Control Application layer Controller parameters: performance index, sample period, controller design, etc. Network layer Routing, flow control, etc. MAC layer Bandwidth sharing through Medium Access Physical layer Modulation, coding, etc. Network design tradeoffs (throughput, delay, loss) implicit in the control performance index

Multiple System Example Inverted Pendulum on a cart. Two identical systems share the network. Different weight matrices in the objective function. q q Actuator Actuator disturbance disturbance Cart Cart u Control force x (position) x (position) Discrete Time Controller Discrete Time Controller

Link Layer Design Tradeoffs (modulation, coding) Uncoded BPSK is optimal!

Iterative Cross Layer Design Example

Optimal vs. Heuristic Controller Codebook Modulation Heuristic Optimal (7,7) BPSK 7.03 6.77 (15,15) 5.77 (31,16) 5.84 (31,11) 5.91 QPSK 5.88 5.53 5.72 5.52   

To Cross or not to Cross? With cross-layering there is higher complexity and less insight. Can we get simple solutions or theorems? What asymptotics make sense in this setting? Is separation optimal across some layers? If not, can we consummate the marriage across them? Burning the candle at both ends We have little insight into cross-layer design. Insight lies in theorems, analysis (elegant and dirty), simulations, and real designs.