Ulas C. Kozat Leandros Tassiulas Mibocom’03 Throughput Capacity of Random Ad Hoc Networks with Infrastructure Support Ulas C. Kozat Leandros Tassiulas Mibocom’03 Ying Huang @ CS598Hou

Motivation Throughput capacity with support of infinite capacity infrastructure network Per-source node capacity Compared with random ad hoc network without infrastructure Assumption Random network in a disk # of ad hoc nodes / access point  Upper bounded W bits/sec with fixed transmission range N ad hoc nodes form a connected topology graph  release later Ying Huang @ CS598Hou

Outline System Model Condition Conclusion Strong Connectivity Weak connectivity Partition, reduce transmission power, less interference, more simultaneous transmission Sufficient and necessary conditions of transmission power Constant bit rate cannot be achieved Achievable capacity Conclusion Ying Huang @ CS598Hou

Two-tier Illustration Ying Huang @ CS598Hou

System Model N, K Transmission is fixed Interference Model number of ad hoc nodes and access points lim N∞(N/K)=α Transmission is fixed Can be arbitrarily small as long as connectivity is maintained Weakly or strongly Interference Model Ying Huang @ CS598Hou

Interference Model Protocol model successful transmission  no other transmitters within a distance (1+D)r of the receiver, where r is the distance from the sender to the receiver receiver sender r (1+D)r m j Ying Huang @ CS598Hou

Outline System Model Condition Conclusion Strong Connectivity Weak connectivity Sufficient and necessary conditions of transmission power Constant bit rate cannot be achieved Achievable capacity Conclusion Ying Huang @ CS598Hou

Capacity Improvement with Infrastructure Layer Per node throughput Key: Make h(N, K) = Θ(1) Access to GW is not the bottleneck Strong Connectivity rT>=sqrt(logN/πN) Ying Huang @ CS598Hou

_____W______ (C+1)*{log(N+K)+1} Intuition _____W______ (C+1)*{log(N+K)+1} λ(N,K)>= =W / logN Interference Cell <=C 1. 3. Θ(1) 2. Θ(log(N+K)) Ying Huang @ CS598Hou

The left  task1 and 3 Each Voronoi cell includes at least one access point The number of destination nodes per access point within a Voronoi cell is Θ(1) # Access points in a cell is Θ(log(N+K)) # destination nodes in a cell is O(log(N+K)) Ying Huang @ CS598Hou

Outline System Model Condition Conclusion Strong Connectivity Weak connectivity Sufficient and necessary conditions of transmission power Constant bit rate cannot be achieved Achievable capacity Conclusion Ying Huang @ CS598Hou

Weak Connectivity Each ad hoc node should be connected to at least one access point with high probability K GW ACi(Xi): capture area Aє: disk area є=rT Ying Huang @ CS598Hou

Probability for being connected to AP Lower bound Upper bound 1-P(disconnected)  - ∞ Ying Huang @ CS598Hou

Weak Connectivity Condition Ying Huang @ CS598Hou

Unachievable of Θ(1) πrT2>c3/K  rT>c4/sqrt(N/π) Under weak connectivity, per node transport capacity of Θ(1) cannot be achieved with p=1  Ying Huang @ CS598Hou

Outline System Model Condition Conclusion Strong Connectivity Weak connectivity Sufficient and necessary conditions of transmission power Constant bit rate cannot be achieved Achievable capacity Conclusion Ying Huang @ CS598Hou

Achievability of Capacity Nuances: Non-uniform distribution Aє(N)=g(N)/N, Lim x∞x2 (Aє(N))’=-∞ Upper bound of per node throughput Each cell includes Θ(g(N)) ad hoc nodes. g(N)/N needs to be defined Ying Huang @ CS598Hou

Conclusion Per source throughput in Hybrid Network Weak connectivity Is Θ(sqrt(N/logN)) better than random ad hoc network without infrastructure Gain is from hop reduction Weak connectivity Ying Huang @ CS598Hou