Ali Saidi

Decimation-in-time-frequency FFT algorithm

A new fast Fourier transform algorithm is presented. The decimation-in-time (DIT) and the decimation-in-frequency (DIF) FFT algorithms are combined to introduce a new FFT algorithm, decimation-in-time-frequency (DITF) FFT algorithm, which reduces the number of real multiplications and additions. The DITF FFT algorithm reduces the arithmetic complexity while using the same computational structure as the conventional Cooley-Tukey (CT) FFT algorithm. The algorithm is extended to radix-R FFT as well as the multidimensional FFT algorithm using the vector-radix FFT.<<ETX>>

CROSS-LAYER DESIGN FOR DYNAMIC RESOURCE ALLOCATION IN WIRELESS NETWORKS

In this paper, a novel analytical cross-layer design framework for dynamic resource management of wireless networks is proposed. First, dynamic bandwidth and time resource allocation policies for a single-user under fading channels that maximize capacity are derived. The analysis is then extended to multi-user environments, where the resource allocation is jointly optimized across both physical and data link layers. A closed-form expression of a QoS measure, mean delay in this case, is derived as a function of layer 2 traffic, multiple access contention from other users, and allocated data-rates at the physical layer. This mean delay expression is then used to efficiently allocate physical layer resources. We also study the effects of various contention mitigation policies on network capacity and average latency under optimum resource allocation strategies.

Battery-aware localization in wireless networks

An important mechanism to conserve energy and extend the battery life of low-power wireless networks is to increase the sleep-cycle of the nodes. However, increasing sleep-cycles can have unintended consequences on the application performance. Therefore, it is important to understand the impact of various sleep-cycle parameters on the performance of any given application. In this paper, we analyze the relationship between the sleep-cycle period and the accuracy of the localization application. We show that the sleep-cycle period has an exponential relationship with many of the parameters that impact localization accuracy. In general, such relationships lend themselves especially well to energy efficient system-level design of location dependent applications. Finally, to demonstrate the validity of our analysis, using an IEEE 802.11 based localization system, we present a set of experimental results that show the relationships between the sleep-cycle and the measurements accuracies in the localization systems.

Social Learning Applications in Resource Constrained Networks

Efficient design of social networking applications must take account of two guiding principles: the adaptive processes by which humans learn and spread new information, and the communication and technological constraints that in turn define the boundaries of human social behavior in virtual communities. In this paper, we introduce the concept of social learning in decentralized, resource-constrained networks. We present a mathematical model for spread of information and derive the optimum strategy that minimizes the total cost of learning in cooperative social networks. We then extend our model to allow individuals to limit their cooperative behavior in spreading their knowledge. Our results demonstrate that increased cooperation reduces the overall cost and accelerates the rate of learning new information.

Minimum-cost First-Push-Then-Pull gossip algorithm

In this paper, we study the communication overhead of gossip-based information dissemination algorithms. Among basic variants of gossip algorithm push is most efficient in the early rounds while, in contrast, pull becomes more efficient in the later rounds. Therefore, a cost-efficient gossip algorithm needs to combine the advantages of push and pull algorithms. One possible approach is to begin with push algorithm and then at some point switch to pull algorithm. We analyze the effect of transition round from push to pull on the communication cost of gossip algorithm. We use simple deterministic difference equations to approximately model the message propagation throughout the network for both push and pull algorithms and derive closed form solution for pull model. We then present our First-Push-Then-Pull (FPTP) gossip algorithm and obtain the optimum round to transition from push to pull. We show that, in a fully connected network, normalized communication cost is minimized to approximately a constant (≈2.6 transmissions/message/node) when the transition round is Round(log N). Furthermore, we extend our results to networks with limited connectivity/cooperation and show that although the communication overhead increases moderately as a function of connection probability, it still remains approximately constant. To validate our results we test our algorithm in mobile ad-hoc network (MANET) environment using random-waypoint mobility model and show that the simulation results closely follow our analysis.

Dynamic Resource Allocation with Outage Probability Constraint for Fading Wireless Channels

In this paper we present a new dynamic resource allocation scheme that enables us to trade-off ergodic capacity for outage probability. We first derive optimal bandwidth and time adaptation policies with constant transmit power subject to average bandwidth and time constraints. These policies lead to maximum ergodic capacities with respect to fixed allocation policies. On the other hand, in the context of water-filling algorithms fixed allocation schemes result in minimum average capacity but have the advantage of simultaneously minimizing outage probability. Our proposed resource allocation technique is a convex combination of adaptive and fixed allocation policies where the outage probability is controlled by varying the boundary between fixed and dynamic allocations. We then apply our algorithm to multiple access networks and using numerical results show that our algorithm is effective in managing the trade-off between average capacity and outage probability for both TDMA and FDMA schemes.

Parallel turbo decoder using a low-latency Max-Log-MAP kernel for a VLIW DSP

Turbo-Codes have attracted great interest in digital mobile radio due to their remarkable error correcting capabilities. In this paper, we present an optimized Turbo decoder for wireless systems following recommended CDMA2000 standard (192 time frames with 8 states per frame). This decoder is implemented using TIs latest C64x digital signal processor. To achieve the maximum parallelism in TIs VLIW architecture, we have specially redesigned the trellis computation algorithm to improve the decoder throughput and reduce the number of computation operations required. In particular, our proposed algorithm transforms a number of add/subtract operations to multiplication operations. In this way, previously unused functional units become available and therefore, more parallel instructions can run simultaneously, leading to throughput increase and latency reduction. To our findings, current TIs compiler (Code Composer V1.2) fails to generate the optimized assembly code when our algorithms are directly implemented in C. To this end, we have illustrated optimized resource binding and timing schedules by applying code motion and loop transformation techniques. The optimized Turbo decoder can finish one decoding stage in 18.1 microseconds for a C64x DSP running at 400 MHz.

Design of 2-D filters composed of low-order systems

In this paper, we introduce a new analytic approach to the design and analysis of 2-D systems composed of low-order systems used as building blocks. We construct low-order 2-D system functions whose unit-circle root contours can be approximated by a straight line. As a counterpart to a 1-D system function with zeros not on the unit-circle, we introduce the notion of system functions whose magnitude response has, approximately, a constant minimum along a controllable linear contour and show how to construct them. Using these low-order system functions as building blocks, in both cascade and parallel forms, we present a simple and efficient algorithm for designing frequency selective 2-D IIR filters based on summing 2-D all-pass filters.

Efficient computation of the large DFT and DCT coefficients

An algorithm for computing the large discrete Fourier transform (DFT) coefficients of a correlated data sequence is introduced. A novel formulation of the decimation-in-frequency (DIF) fast Fourier transform (FFT) algorithm is introduced which generalizes the decimation method FFT algorithm. The radix-2 DIF FFT algorithm is modified to introduce an efficient algorithm for computing the DFT coefficients larger than a given threshold. This algorithm significantly reduces the computations associated with the small coefficients. Results are extended to all the DIF-based FFT algorithms as well as the multidimensional FFT and the FFT-based fast discrete cosine transform algorithm.<<ETX>>