for Vehicular Ad Hoc Networks

Transcription

1 Distributed Fair Transmit Power Adjustment for Vehicular Ad Hoc Networks Third Annual IEEE Communications Society Conference on Sensor, Mesh and Ad Hoc Communications and Networks (SECON 06) Reston, VA, September 2006 Marc Torrent-Moreno, Paolo Santi *, Hannes Hartenstein University of Karlsruhe, Germany * Consiglio Nazionale delle Ricerche, Italy Marc Torrent-Moreno

2 Which Problem are we looking at? Simple scenario: Simple question: - Which transmission power should I use? Not so simple constraint: - Main gaol of VANETs: Safety Marc Torrent-Moreno - 2

3 Structure of this talk» Introduction / Motivation» TxPower and Fairness» Problem Statement» Centralized solution and its limitations» Distributed solution and its challenges» Simulation results» Summary, conclusions and outlook Marc Torrent-Moreno - 3

4 How safety is addressed in VANETs: Awarness Beaconing messages: Periodic (few/s) 1-Hop Broadcast Advertise status information (Position, speed ) Provide awareness of a vehicle s close surroundings at all times Information allows detection of unsafe situation Event-driven messages: Upon emergency detection Mutli-hop geo-broadcast Disseminate information about an unsafe situation Provide awareness of a hazard to all cars up to large distances Marc Torrent-Moreno - 4

5 Inter-Vehicle Communication: Facts» p is a variant of a with 10MHz channels (half data-rates)» One channel, Control Channel, for safety information exchange» CSMA/CA is an asynchronous approach, not able to manage bandwidth efficiently» Safety messages are mainly broadcast, hidden terminal problem» Potential fast moving heavy vehicular traffic conditions high channel load dynamic topology high number of packet collisions low safety Marc Torrent-Moreno - 5

6 Rationale How can we control the channel load?» Event-driven messages Relevant for close and further distances Fast dissemination is critical» Beacons Relevant for close in distances Information updated periodically Efficient strategy Amount of load resulting from beaconing should be limited How?» Adjusting packet generation rate» Adjusting transmission power Marc Torrent-Moreno - 6

7 Which power adjustment strategy addresses safety?» Capacity and connectivity are NOT the goals of VANETs» TxPower adjustment should be build upon the concept of Fairness: Every vehicle needs a good estimation of the state (position, speed) of all vehicles (with no exception) in its close surrounding A vehicle without a fair assignment of the resources can became a danger itself Max-Min Fairness D-FPAV Distributed Fair Tx Power Adjustment for Vehicular Environments Marc Torrent-Moreno - 7

8 Problem statement: BL(PA) MaxBL BL=2 BL=3 PA: Power Assignment Pi Pmax: Transmission Power where Pi [0,1] CS range PA(1) PA(2) PA(3) PA(n) BL: Beaconing Load» Goal: Beaconing Max-Min Tx Power (BMMTxP) For simplicity, all vehicles are configured with the same: Intended TxPower Packet Generation Rate Beacon size Marc Torrent-Moreno - 8

9 1st step, solve the problem globally: FPAV [ACM-VANET05] Input: A set of nodes { u u n} N =,..., 1 Output: a power assignment PA wich is an (ε Pmax-approximation of an) optimal solution to BMMTxP u while( BL( PA) MBL) end u i i N, set PA( i) = 0 u i N, PA( i) = PA do ( i) + ε while N, PA( i) = PA( i) ε *Inspired in Water Filling [Bersktas87] Marc Torrent-Moreno - 9

10 BL 2 1 FPAV Visualization example MaxBL = 2 Marc Torrent-Moreno - 10

11 FPAV Characteristics and Limitations» FPAV would achieve the goals identified: Limits the wireless channel load resulting from periodic messages Achieves Max-Min FAIRNESS» However, FPAV is a centralized algorithm, not practical in a real vehicular scenario» Next step: Distributed strategy 2 nd Step: D-FPAV fully distributed, localized and asynchronous implementation Marc Torrent-Moreno - 11

12 D-FPAV Intuition 1. Which nodes can a transmission directly affect? 2. Which transmissions can affect a node? Each node needs to know how all cars in his CS see the medium CSmax Marc Torrent-Moreno - 12

13 D-FPAV, Algorithm Input: Status of all nodes in CS max (i) Output: A power setting PA(i) for node u i such that PA is optimal solution of BMMTxP Step 1. Compute common maximum tx power P(i) of all nodes, so BL < MBL using FPAV Step 2a. Broadcast P(i) to all nodes in CS max Step 2b. Receive messages with P(j) from nodes j in CS max Step 3. Compute final power level as the minimum of the received power levels and P(i) Proven formally to achieve fairness Marc Torrent-Moreno - 13

14 Estimation of status information How do I get the information from neighbors inside CS?» Two main options to obtain data from nodes outside CR: Increase TxPower of few beacons with own D-FPAV data Piggyback status + D-FPAV data from neighbors in few beacons» A tradeoff beween accuracy and overhead exists» How much Fairness cost? (How often must be this information piggybacked?) First approach, strategies simulated: Every beacon with adjusted TxPower Two out of ten beacons with adjusted TxPower Two out of ten beacons with maximum TxPower One out of ten beacons with adjusted TxPower One out of ten beacons with maximum TxPower Marc Torrent-Moreno - 14

15 Simulation Analysis a) HWGui Realistic highway traces from Fleetnet/NoW project b) ns-2.28 Bug-fixes MAC layer Adjustment of PHY to p 3 Radio propagation models Marc Torrent-Moreno - 15

16 D-FPAV, Channel Busy Time Nakagami propagation model Reduce the load resulting from beaconing Fair distribution of the load on the channel Marc Torrent-Moreno - 16

17 D-FPAV, Probability of Reception Nakagami propagation model Increase of reception rates of event-driven messages at all distances Maintain good reception rates at distances close to the sender for beaconing messages Marc Torrent-Moreno - 17

18 Summary and Conclusions» Due to the limited channel capacity of VANETs we propose to restrict the wireless channel load resulting from periodic messages» Due to the SAFETY nature of VANETs, we propose to built channel congestion avoidance strategy built upon FAIRNESS» D-FPAV is a totally distributed, localized and asynchronous strategy to adjust the periodic messages TxPower» With D-FPAV: The load on the channel is controlled: beaconing_load MaxBL The bandwidth resources are assigned in a fair manner Event-driven messages reception probabilities are increased at all distances, improving reliability and information dissemination speed Beacons maintain good reception rates for vehicles located in the close in distances Marc Torrent-Moreno - 18

19 Outlook We are working on...» With respect to D-FPAV: Adjust MaxBL, balance the relevance of beacons and event-driven messages Find estimation of the optimal point of operation (tradeoff overhead-accuracy)» Strategy to disseminate information quickly and reliable on a loaded channel Marc Torrent-Moreno - 19

20 Any Questions? Marc Torrent-Moreno Web: Marc Torrent-Moreno