Link/Network Layer: MIMO, Cognitive Radio; Energy Management of Radio Resource Control (RRC) Y. Richard Yang 11/15/2012

2 Outline r Admin. and recap r Improve mesh capacity m Reduce L (infrastructure “blackholes”, mobility for delay tolerant networks) m MIMO: Use multiple antennas m Cognitive radio: use more spectrum r Radio resource management for energy management of mobile devices

Admin. r Project meeting slots to be posted on classesv2 3

4 Recap: Constraints in Capacity Analysis  transmission successful if there are no other transmitters within a distance (1+  )r of the receiver Interference constraint r a single half-duplex transceiver at each node Radio interface constraint receiver sender r (1+  )r

5 Recap: Capacity Bound Note: Let L be the average (direct-line) distance for all T end-to-end bits. rate*distance capacity:

6 Improving Wireless Mesh Capacity  transmission successful if there are no other transmitters within a distance (1+  )r of the receiver Interference constraint r a single half-duplex transceiver at each node Radio interface constraint rate*distance capacity: Reduce LIncrease W Approx. optimal Multiple transceivers Reduce interf. area

7 Outline r Admin. and recap r Improve mesh capacity m Reduce L (infrastructure “blackholes”, mobility for delay tolerant networks) m MIMO: Use multiple antennas

Multiple Input Multiple Output (MIMO) r 4x4 MIMO m r LTE r Kindle Fire HD 8

MIMO Basics Solve two variables from two equations.

Using MIMO for more Concurrency: Motivation 10 Assume tx1 is sending to rx1 Can tx2 transmit in using carrier sensing? No Transmission in current n

MIMO Benefit: Concurrency using Interference Nulling tx2: for every symbol q, transmits q on first antenna and aq on second antenna. 11 interference at rx1: if tx2 picks NO interference at rx1.

Problem - rx2 hears p from tx1 - Can rx2 decode? 12

Decoding at rx2: Observation - for different symbols p from tx1, the received signal at rx2 moves along a 1-d vector 13 - rx2 can estimate channels h12, h13 from preamble Perp. Of tx1 space

Decoding at rx2: Removing tx1 signal by Projection - rx2 projects received signal orthogonal to 14 projection space

Decoding at rx2: Projection Details - rx2 picks w2 and w3: to compute 15 projection space

Decoding at rx2: Projection Details 16 => Summary: MIMO allows concurrency w/ interference nulling.

Problem of Only Nulling 17 Assume both tx1 and tx2 are transmitting. If only nulling, tx3 cannot transmit nulling

Solution: MIMO using Interference Alignment 18 Assume both tx1 and tx2 are transmitting. Key idea: rx2 ignores interference from tx1 by projection. If tx3 aligns tx3 -> rx2 interference along the same direction as that of tx1 -> rx2, then rx2 can remove it too. nulling

MIMO with Nulling and Alignment 19 tx3 picks  ’,  ’,  ’ rx2 sees: Because rx2 projects to orthogonal to, no interference from tx3 to rx2

20 Outline r Admin. and recap r Improve mesh capacity m Reduce L (infrastructure “blackholes”, mobility for delay tolerant networks) m MIMO: Use multiple antennas m Cognitive radio: use unlicensed spectrum

Spectrum Allocation Chart 21

Unlicensed Spectrum r Opportunity: unlicensed spectrum is large and has low utilization m US unlicensed freq: G M G G (200 mw) (1 w) (4w) 22

Problem of Using Unlicenced r Unlicensed spectrum may have occupants and is fragmented r Requirement: Coexistence with dynamic and unknown narrowband devices in the unlicensed spectrum 23 Zigbee a Others Unlicensed Spectrum

Zigbee a Others Unlicensed Spectrum 1.Operate below noise-level Limits range Wideband Existing Solutions

Zigbee a Others Unlicensed Spectrum 1.Operate below noise-level Limits range 2. Pick a contiguous unoccupied band Limits throughput Wideband Existing Solutions

Zigbee a Others Unlicensed Spectrum 1.Operate below noise-level Limits range 2. Pick a contiguous unoccupied band Limits throughput Wideband Sacrifice Throughput or Range! Existing Solutions

Zigbee a Others Unlicensed Spectrum Cognition: Detect unoccupied bands Aggregation: Weave all unoccupied bands into one link Wideband Swift: Cognitive Aggregation

Research Issues r How to detect available frequency bands? r How to operate across chunks of non- contiguous frequencies? r How do sender and receiver establish communication when their perceived available frequency bands differ?

Aggregating Non-Contiguous Bands Leverage OFDM Divides frequency band into multiple sub-bands that can be treated independently Transmitter: Puts power and data only in OFDM bands not occupied by narrowband devices Receiver: Extracts data only from OFDM bands used by transmitter Frequency band

Cognition: How to detect occupied bands? Unlicensed  Can’t assume known narrowband devices Typical solution: Power threshold Ideal Threshold Narrowband Power in dBm Baseband Frequencies (MHz) Faraway

Cognition: How to detect occupied bands? Unlicensed  Can’t assume known narrowband devices Typical solution: Power threshold Ideal Threshold Narrowband Power in dBm Baseband Frequencies (MHz) Faraway Problem: No Single Threshold Works Across All Locations

Ideal Threshold Narrowband Power in dBm Baseband Frequencies (MHz) Faraway Nearby Unlicensed  Can’t assume known narrowband devices Typical solution: Power threshold Cognition: How to detect occupied bands?

Adaptive Sensing Intuitively: Poke the narrowband device, putting power in ambiguous bands If the narrowband device reacts, back away Reasonable for unlicensed spectrum, which operates as best-effort Unlicensed devices typically react to interference Carrier sense in , TCP backoff, etc.

Continuously sense the medium when not sending a packet Detect appearance of narrowband device when narrowband power exceeds noise level Detect reaction from changes in narrowband power profile Adaptive Sensing: Alg

Narrowband ReactionDetection Metric Carrier Sense (e.g.,802.11): Will not transmit when sensing a SWIFT packet Probability of narrowband power immediately after a SWIFT packet Back-off (e.g., TCP, MAC): Will send less often Inter-arrivals of narrowband power Autorate: Will use lower modulation, increasing packet size Duration of narrowband power Look for statistically significant change in metric using standard tests (e.g. t-test)

Adaptive Sensing in Action r Start with a conservative choice of bands r Keep tightening as long as narrowband is unaffected Conservative Threshold

Wideband Start with a conservative choice of bands Keep tightening as long as narrowband is unaffected Adaptive Sensing in Action

Wideband Time Metric Estimate Normal Behavior Adaptive Sensing in Action

Wideband Time Metric Tighten Sense Test: Same as Normal Adaptive Sensing in Action

Wideband Time Metric Tighten Sense Test: Different from Normal Adaptive Sensing in Action

Wideband Time Metric Loosen Sense Test: Same as Normal Adaptive Sensing in Action

Wideband Throughput and Range Baseline that operates below the noise of

Wideband Throughput and Range

Other Work Cognitive Radios , KNOWS, CORVUS, DIMSUMNet etc. Wideband systems Intel, Chandrakasan et al., Mishra et al., Sodini et al.

45 Outline r Admin. and recap r Improve mesh capacity r Radio resource management for energy management

46 Recall: GSM Logical Channels and Request r Many link layers use a hybrid approach m Mobile device uses random access to request radio resource m The device holds the radio resource during a session BTS MS RACH (request signaling channel) AGCH (assign signaling channel) SDCCH (request call setup) SDCCH (assign TCH) SDCCH message exchange Communication call setup from an MS

47 Radio Resource Control Setup for Data in 3G RRC connection setup: ~ 1sec Radio Bearer Setup: ~ 1 sec + Figure source: HSDPA/HSUPA for UMTS: High Speed Radio Access for Mobile Communications. John Wiley and Sons, Inc., Source: Erran Li.

48 RRC State Management in UMTS r Given the large overhead to set up radio resources, UMTS implements RRC state machine on mobile devices for data connection ChannelRadio Power IDLENot allocated Almost zero CELL_FAC H Shared, Low Speed Low CELL_DCHDedicated, High Speed High Courtesy: Erran Li.

49 RRC of a Large Commercial 3G Net Promo Delay: 2 Sec DCH Tail: 5 sec FACH Tail: 12 sec DCH: High Power State (high throughput and power consumption) FACH: Low Power State (low throughput and power consumption) IDLE: No radio resource allocated Tail Time: waiting inactivity timers to expire

50 RRC Effects on Device/Network FACH and DCH Wasted Radio Energy34% Wasted Channel Occupation Time33%

51 Case Study: Pandora Streaming Problem: High resource overhead of periodic audience measurements (every 1 min) Recommendation: Delay transfers and batch them with delay-sensitive transfers Problem: High resource overhead of periodic audience measurements (every 1 min) Recommendation: Delay transfers and batch them with delay-sensitive transfers

52 Case Study: Fox News Problem: Scattered bursts due to scrolling Recommendation: Group transfers of small thumbnail images in one burst Problem: Scattered bursts due to scrolling Recommendation: Group transfers of small thumbnail images in one burst

53 Case Study: BBC News Scattered bursts of delayed FIN/RST Packets Problem: Scattered bursts of delayed FIN/RST packets Recommendation: Close a connection immediately if possible, or within tail time Problem: Scattered bursts of delayed FIN/RST packets Recommendation: Close a connection immediately if possible, or within tail time

54 Case Study: Google Search Search three key words. ARO computes energy consumption for three phases I: Input phase S: Search phase T: Tail Phase UL Packets DL Packets Bursts RRC States Usr Input Problem: High resource overhead of query suggestions and instant search Recommendation: Balance between functionality and resource when battery is low Problem: High resource overhead of query suggestions and instant search Recommendation: Balance between functionality and resource when battery is low

55 RRC State Transitions in LTE

56 RRC State Transitions in LTE RRC_IDLE No radio resource allocated Low power state: 11.36mW average power Promotion delay from RRC_IDLE to RRC_CONNECTED: 260ms

RRC state transitions in LTE RRC_CONNECTED Radio resource allocated Power state is a function of data rate: 1060mW is the base power consumption Up to 3300mW transmitting at full speed 57 Cellular Networks and Mobile Computing (COMS ) Courtesy: Junxian Huang et al.

RRC state transitions in LTE Continuous Reception Send/receive a packet Promote to RRC_CONNECTED Send/receive a packet Promote to RRC_CONNECTED Reset Ttail 58 Cellular Networks and Mobile Computing (COMS ) Courtesy: Junxian Huang et al.

RRC state transitions in LTE Ttail stops Demote to RRC_IDLE DRX Ttail expires 59 Cellular Networks and Mobile Computing (COMS ) Courtesy: Junxian Huang et al.

60 Summary r App developers may not be aware of interactions with underlying network radio resource management r A good topic to think about as a part of your project