1 / 14

Low complexity and distributed energy minimization

Low complexity and distributed energy minimization . L. Lin, X. Lin and N. B. Shroff Presented by: Srikanth Hariharan. Introduction. Goal: To minimize total power consumption in a multi-hop wireless network subject to a given load.

tamra
Download Presentation

Low complexity and distributed energy minimization

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Low complexity and distributed energy minimization L. Lin, X. Lin and N. B. Shroff Presented by: SrikanthHariharan

  2. Introduction • Goal: To minimize total power consumption in a multi-hop wireless network subject to a given load. • Total power consumption can be reduced by jointly optimizing power control, scheduling and routing. • Known optimal solution is centralized with high computational complexity. • Approximate optimal solution is also centralized and it yields a 3-approximation ratio. • Here, a low complexity and distributed (2+ε) - approximate solution is developed.

  3. Notations and Model • G(V,E) – Network with nodes V and links E. • No(v) – Outgoing links of node v. • Ni(v) – Incoming links of node v. • Node-exclusive interference model and time slotted system. • h(Re) – Power to support a data rate Re of link e – assumed to be convex, non-decreasing with h(0)=0. • D – set of destinations. • Tvd – Long-term average data rate of the flow that needs to be supported from source node v to destination d. • fed – Average amount of data rate on link e allocated for destination d.

  4. Problem Formulation Problem: Scheduling component in (*) has a high computational complexity.

  5. Approximation • For all time-slots mwhen link e is activated, Re(m) is independent of m in any power-optimal scheme. • So, the objective function of (*) is now • From existing results on low – complexity scheduling, we have and if , where ηis arbitrarily small, then a maximal schedule can be computed such that each link is activated for fraction of time-slots. • Replace scheduling constraints by • β = 1 – Lower bound. • β = ½ - η– Upper bound on optimum.

  6. Handling Non-convexity • Objective function and constraints non-convex. • Solution – Do change of variable • te – Fraction of time-slots link e is activated. • te = 0 => fed = 0 for all d. • Then, the long term average power consumption from link e can be expressed as

  7. Convex optimization problem

  8. Lagrange Duality and Energy Minimization Algorithm Strong duality holds.

  9. Interpretation • Power cost – • Scheduling cost – • Utility – • can be interpreted as an approximation of the scaled version of the queue length at node v for destination d if constant step sizes are used. le is minimum when all the data rates are assigned to the destination with the maximum positive backlog difference.

  10. Interpretation • What is the optimal te? • When is te = 0? • All backlog differences are negative => Utility does not increase by transporting data to next hop on this link. • Energy saving and/or the need to gain more flexibility in scheduling is more important than the utility of transporting data to next hop.

  11. Components • Routing: Choose the flow with maximum positive backlog difference. • Power control: Choose Re to minimize le(Re). • Link Assignment: Choose te to minimize tele(Re) – This determines the amount of time link e should be on. • Maximal Scheduling: Need to determine exact time slots for link e given the fraction of up-time • Distributed algorithms exist for the node-exclusive interference model when .

  12. Convergence Results • Constant step sizes: When step sizes are small, • dual variables converge to within a small neighborhood of the optimal dual solution • primal variables may not converge, however, the long-term average of the resultant power consumption is arbitrarily close to the optimal power consumption. • Diminishing step sizes: For step sizes that satisfy , the algorithm converges to the optimal value. • If the step sizes are chosen as , if f and t converge, then . • Power efficiency ratio in an AWGN channel is at most (2+ε) that of the optimal value with β=1.

  13. Numerical Results t= 4000: σ(1,7) (gain) decreases. t=8000: Data rate of flow 2 decreases to 250 kbps.

  14. Conclusion • Joint power control, link scheduling and routing algorithm proposed to minimize energy. • Distributed algorithm with low computational complexity. • Power efficiency ratio tighter than existing solution. • Solution adapts to changes in channel conditions and flow rates. • Methodology can be extended to a general interference model.

More Related