Cooperation in Random Access Wireless Networks

Transcription

1 Cooperation in Random Access Wireless Networks Presented by: Frank Prihoda Advisor: Dr. Athina Petropulu Communications and Signal Processing Laboratory (CSPL) Electrical and Computer Engineering Department Drexel University

2 Outline of Discussion Introduction/Background Random Access Methods ALOHA NDMA Proposed Cooperation Technique ALLIANCES Current Work Relay Selection Conclusion Questions

3 Background Multimedia sources require diverse bandwidth requirements. Fixed Allocation Schemes are insufficient for variable-bitrate applications. Simple random access schemes such as ALOHA suffer from limited throughput. The focus of previous multi-user research was to avoid collisions.

4 Random Access Methods ALOHA In random access methods such as ALOHA, each user transmits its packet when it has a packet to transmit. As the traffic load in the network increases, packets at the receiver will collide. In the ALOHA scheme, collided packets are discarded. C onsider a slotted-system User 1 slot n-4 slot n-3 slot n-2 slot n-1 Collision slot n User 2 User 3 User 4 User 5 User 6

5 Random Access Methods By introducing diversity to the wireless network approach, an advantage is provided over the original ALOHA scheme. Multi-user detection and source separation can be used at the receiver to detect collisions (and the collision order). In a recent method (NDMA) it was proposed that in a K fold collision, the packets from the collision time slot are saved in buffer, and using retransmissions during the following K-1 time slots, the originally collided packets can be resolved.

6 Random Access Methods NDMA Consider the three fold collision in time slot n (previous slide) 3 Fold Collision

7 Random Access Methods NDMA

8 Random Access Methods NDMA Collision can be resolved

9 Random Access Methods NDMA NDMA exploits time diversity in the wireless network through retransmission of collided packets. Usually, channel coefficients are correlated over adjacent time slots. If a certain channel is in deep fading, it will continue to be in deep fade during the following time slots. Deep Fading will reduce the rank of the channel coefficient matrix A. NDMA provides higher throughput than ALOHA, however, it is best suited for fast varying channels.

10 Random Access Methods ALLIANCES ALLow Improved Access in the Network via Cooperation and Energy Savings In addition to time diversity, we introduce spatial diversity through user cooperation. The additional diversity introduced provides us with a higher throughput without additional hardware or bandwidth expansion. View the wireless network as a spatially distributed antenna, with all elements linked via the wireless channel. A savings in network power efficiency can be observed.

11 Random Access Methods ALLIANCES Assumptions: Small-scale multi-user slotted system Transmitters are synchronized and in half duplex mode Each node is equipped with one antenna Nodes will receive feedback from base station specifying if the transmission was successful Only non-regenerative relays will be used

12 Random Access Methods ALLIANCES When a collision is detected, the system will enter a cooperative transmission epoch (CTE). A control bit will be broadcast to all nodes indicating the start of CTE. Initially, the CTE will be of length K-1. If necessary, the base station will increase this length. Selection of relays during CTE Randomly SNR User Energy (Available Battery Power)

13 Random Access Methods ALLIANCES Consider the same collision during time slot n a 10 (n) y 0 (n) received at base station 3 Fold Collision a 20 (n) a 30 (n) Signal at BS/AP y 0 (n) = a 10 (n)x 1 (n) + a 20 (n)x 2 (n) + a 30 (n)x 3 (n) Signal at i th relay node y i (n) = a 1i (n)x 1 (n) + a 2i (n)x 2 (n) + a 3i (n)x 3 (n)

14 Random Access Methods ALLIANCES In the first time slot of the CTE, a chosen relay bounces the signal it received during the collision time slot a 40 (n+1) y 0 (n) received at base station Signals stored at BS/AP time slot n y 0 (n) = a 10 (n)x 1 (n) + a 20 (n)x 2 (n) + a 30 (n)x 3 (n) time slot n+1 y 0 (n+1) = a 40 (n)y 4 (n)

15 Random Access Methods ALLIANCES For the 3 Fold Collision example, the originally collided packets are resolved in time slot n+2 Collision can be resolved y 0 (n) received at base station a 30 (n+2) Y = AX + W Signals stored at BS/AP time slot n y 0 (n) = a 10 (n)x 1 (n) + a 20 (n)x 2 (n) + a 30 (n)x 3 (n) time slot n+1 time slot n+2 y 0 (n+1) = a 40 (n)y 4 (n) y 0 (n+2) = a 30 (n+2)x 3 (n) Note: In this situation, we show the special case where the selected relay is a source node. In this case (as shown), the user will simply retransmit its original packet.

16 Simulation We consider a wireless system with 32 users Orthogonal IDs were added to 4 QAM packets, with N = 424 bits. Symbols were recovered using a zero forcing function at the receiver If the packets were successfully detected at the receiver with an error probability.02, the transmission was considered successful Jake s model was used to simulate the fading channel. A bit rate of 256 kbps and Doppler frequency shift of 52 Hz. With a carrier frequency of 5.2GHz, this corresponds to a user speed of 3 m/s. The following results are simulations of the network to handle an aggregated traffic load (both new traffic and traffic from retransmissions).

17 Simulated Results

18 Application Advantage of Cooperation: Consider a military scenario, where we view the network nodes as soldiers carrying antennas. With cooperation, we can enable uninterrupted, high-bit rate (multimedia e.g. voice, data, video) communication, EVEN if the some of the users face poor channel conditions (i.e. poor SNR, no LOS etc.).

19 Selection of Relays In the originally proposed ALLIANCES scheme, relays were selected at random. For example, R i = mod(n + i, J) + 1. For this, J was the set of all users. To see an improved throughput, we can consider using different relay sets; with respect to set size and node choice. Relay selection may provide us with additional control overhead. This overhead can be justified, if an improvement in throughput is observed.

20 Selection of Relays Selection of relay via user location. Relay sets of size J/2 and K-1 Selection of relay from the joint set of user s location and user s available battery Relay set of size J/2 Compare the above with the original selection technique in the proposed method

21 Selection of Relays

22 Simulation

23 Conclusions Spatial diversity provided by the ALLIANCES scheme is robust in a wireless environment, with throughput improvement over both NDMA and ALOHA. ALLIANCES is suitable for a wireless network in which nodes are willing to cooperate. In addition to the cooperation advantage, ALLIANCES is advantageous for bursty sources. ALLIANCES only requires one node to transmit during the CTE, thus, when compared to NDMA, it provides us with a more energy efficient way of resolving collisions. Throughput performance can be increased with different relay selection techniques. However, we must make note that these additional techniques will add control overhead to the protocol.

24 Questions?