Uncoordinated Multiple Access for Vehicular Networks

Transcription

1 Uncoordinated Multiple Access for Vehicular Networks Mikhail Ivanov, Fredrik Brännström, Alexandre Graell i Amat, Petar Popovski Chalmers University of Technology, Gothenburg, Sweden Aalborg University, Aalborg, Denmark WWVC, Halmstad, November 11, 2014

2 Motivation V2X communications Numerous applications, e.g., cooperative intelligent transportation systems, automated driving, traffic safety, traffic efficiency, etc. State-of-the-art ad hoc V2X communications p (slightly modified Wi-Fi) using carrier sense medium access (CSMA) Main challenges difficult propagation channels potentially fast changing network topologies very strict delay requirements very strict reliability requirements all-to-all broadcast Uncoordinated Multiple Access for Vehicular Networks Mikhail Ivanov, Fredrik Brännström, Alexandre Graell i Amat, Petar Popovski 1 / 14

3 Outline 1. Coded Slotted ALOHA: Unicast 2. Error Floor Analysis for Finite Frame Lengths 3. Coded Slotted ALOHA: Broadcast 4. Coded Slotted ALOHA vs Carrier Sense Multiple Access 5. Conclusions and Future Work Uncoordinated Multiple Access for Vehicular Networks Mikhail Ivanov, Fredrik Brännström, Alexandre Graell i Amat, Petar Popovski 2 / 14

4 Encoding Assumptions Time is divided into frames each of N slots of equal duration. At the beginning of each frame every user has a message to transmit. The message size matches the slot duration. Distribution: Λ(x) = 0.75x x 3 1 (2) 2 (2) 3 (3)... M N t Uncoordinated Multiple Access for Vehicular Networks Mikhail Ivanov, Fredrik Brännström, Alexandre Graell i Amat, Petar Popovski 3 / 14

5 Decoding Assumptions The base station records the entire frame. A packet can be reliably decoded if no collision occurs. Each packet contains pointers to its replicas. The replicas of a received packet can be perfectly removed from the signal M N t Uncoordinated Multiple Access for Vehicular Networks Mikhail Ivanov, Fredrik Brännström, Alexandre Graell i Amat, Petar Popovski 4 / 14

6 Parameters number of slots N channel load G = M N packet loss rate PLR = Performance Measure number of users M # resolved users throughput Th = N M # resolved users M = 1 Th G 1.0 Throughput 10 0 Packet loss rate MAC is usually interested in Throughput PLR ǫ 0.2 We are interested in G [user/slot] G ǫ G [user/slot] Uncoordinated Multiple Access for Vehicular Networks Mikhail Ivanov, Fredrik Brännström, Alexandre Graell i Amat, Petar Popovski 5 / 14

7 CSA and Codes on Graphs Identical to LDPC decoder on binary erasure channel! Similarities believe propagation on erasure channel density evolution (EXIT chart) threshold phenomenon Differences no channel graph is not known a priori negative code rate Uncoordinated Multiple Access for Vehicular Networks Mikhail Ivanov, Fredrik Brännström, Alexandre Graell i Amat, Petar Popovski 6 / 14

8 Finite Frame Length Distribution Λ(x) = 0.15x x x 8 Frame length N = PLR 10 3 tx 2 DE 10 4 average 10 5 tx 3 tx G [user/slot] Uncoordinated Multiple Access for Vehicular Networks Mikhail Ivanov, Fredrik Brännström, Alexandre Graell i Amat, Petar Popovski 7 / 14

9 Stopping Sets Uncoordinated Multiple Access for Vehicular Networks Mikhail Ivanov, Fredrik Brännström, Alexandre Graell i Amat, Petar Popovski 8 / 14

10 Error Floor Distribution Λ(x) = 0.15x x x 8 Frame length N = PLR tx 2 average tx G [user/slot] tx 8 Uncoordinated Multiple Access for Vehicular Networks Mikhail Ivanov, Fredrik Brännström, Alexandre Graell i Amat, Petar Popovski 9 / 14

11 All-to-All Broadcast Nothing is decodable! Notes with different degrees have different decoding capabilities. Removing a finite number of the columns does not affect the asymptotic analysis. Removing slots can be seen as changing the distribution. Degree-1 users are present in the new distribution. Uncoordinated Multiple Access for Vehicular Networks Mikhail Ivanov, Fredrik Brännström, Alexandre Graell i Amat, Petar Popovski 10 / 14

12 Stopping Sets Additional Stopping Sets Uncoordinated Multiple Access for Vehicular Networks Mikhail Ivanov, Fredrik Brännström, Alexandre Graell i Amat, Petar Popovski 11 / 14

13 10 2 Simulation Results Distribution Λ(x) = 0.15x x x 8 Frame length N = 200 PLR average rx 3 rx 2 rx G [user/slot] Uncoordinated Multiple Access for Vehicular Networks Mikhail Ivanov, Fredrik Brännström, Alexandre Graell i Amat, Petar Popovski 12 / 14

14 CSA vs CSMA Distribution Λ(x) = 0.86x x N = 172 N = PLR G [user/slot] Uncoordinated Multiple Access for Vehicular Networks Mikhail Ivanov, Fredrik Brännström, Alexandre Graell i Amat, Petar Popovski 13 / 14

15 Conclusions Conclusions and Future Work Finite frame length analysis is reasonably accurate Finite frame length is needed for controlling latency All-to-all broadcast is different from traditional MAC both in terms of performance metrics and analysis CSA seems to significantly outperform CSMA Future work Design of good distributions for all-to-all broadcast Error floor reduction How to utilize (priorities?) or remove unequal error protection SINR model for interference cancellation Introduce frameless structure Thank you! Uncoordinated Multiple Access for Vehicular Networks Mikhail Ivanov, Fredrik Brännström, Alexandre Graell i Amat, Petar Popovski 14 / 14