Research interestsMy research interests lies in the development and application of models and analytical tools for performance analysis and optimization of engineering systems. Specifically, I am interested in the development and application of queueing models and analytical tools for communication and computer networks. However, I am also interested in several other problems. Please read on to find what all I am working on as well as related publications. Research workStability and delay performance for wireless systems with reconfiguration delay We consider wireless downlinks where the base station switches between different users in order to transmit data intended for the respective users. When switching between users, the logical links between the user and the base station need to be configured. For example, the user’s state may need to be changed from an idle to an active state. This incurs a reconfiguration delay, which is the delay between the time at which the base station scheduler decides to serve the user and the time at which the actual data transmission starts. Motivated by such scenarios, we consider the stability region and throughput optimal scheduling for a wireless downlink model with random connectivity over time and reconfiguration delay. We have also looked at the problem of minimizing average delay for wireless systems with reconfiguration delay when the channel connectivity is correlated
Stability and delay analysis for delay tolerant networks with random packet arrivals Our research objective is to theoretically analyze the stability of queues and average delay of messages in a delay tolerant network. We consider a simple delay tolerant network model with the source spray and wait routing protocol without feedback for our initial analysis. This is work in progress. Some approximations to the average queueing delay has been obtained by interpreting the system as a M^{K}/G/N system. Simulation code for this work is available here.
Bayesian quickest detection for active sensors Active sensors use radar, ultrasound, or optical sensing to control the information content in the sequentially collected noisy data. Using such active sensing to increase the extent of collected information, it is possible to quickly detect changes in the environment. However, increasing the extent of information contained in the collected data would lead to a cost or energy expenditure, where the cost would naturally increase as the extent of contained information increases. For such a sensor, our objective is to design control policies that optimally trade off performance metrics such as detection delay, average cost expended in control, and probability of false alarm. We consider this problem in the setting of Bayesian quickest change detection. Our preliminary results can be found in the following paper.
Optimal addition of links to a communication network The quality of service offered by a communication network to its users can be improved by the addition of communication links between appropriate nodes in the network. We pose this link addition problem as an optimal edge addition problem in graphs. Some of our preliminary results are given in the following paper, where we have considered a simple linear graph.
Power delay tradeoff for dynamic data compression and transmission over a wireless channel In power constrained sensor networks and satellite telemetry systems there arises scenarios where multiple sensors produce data which is correlated. We consider a communication scenario where a power-limited transmitter aggregates correlated data from different sources for transmission to a sink over a wireless fading channel. In order to reduce the average queueing delay, the transmitter could compress the data before transmission. We note that it is not obvious that compressing the data to the maximum extent possible leads to the best possible tradeoff. Prior work has found that although wireless transmission of a bit can require energy over a 1000 times more than a single 32-bit computation, compression using typical algorithms (LZW) may actually lead to larger power expenditure, due to power expensive memory accesses. So both data compression as well as wireless transmission expends comparable amounts of power which is a precious resource for the transmitter. We consider the problem of designing joint compression and transmission policies for the transmitter which can optimally tradeoff the average queueing delay of the transmitted data with the average total power of compression and transmission. Our preliminary results are reported in this work.
We studied the tradeoff of average delay with average service cost and average utility, for single server queueing models without and with admission control. We considered continuous time and discrete time queueing models with a random environment. The tradeoff problem is studied for the class of monotone scheduling policies, i.e., scheduling policies for which the service rate is a non-decreasing function of the queue length for each environment state. The continuous time and discrete time queueing models that we considered are motivated by cross-layer models for point-to-point links with random packet arrivals and fading at slow and fast time scales. Our objective is motivated by the need to optimally tradeoff the average delay of the packets (a network layer performance measure), with the average service cost of transmitting the packets, e.g. the average power required for transmission (a physical layer performance measure), under a lower bound constraint on the average throughput. We also considered the problem of optimally trading off the average delay and average error rate of randomly arriving message symbols which are transmitted over a noisy point-to-point link. It is intuitive that to keep a queue stable under a lower bound constraint on the average utility, a minimum number of customers have to be served per unit time. This in turn implies that queue stability requires a minimum average service cost expenditure. In this thesis, we obtain primarily an asymptotic characterization of the minimum average delay for monotone policies, subject to an upper bound constraint on the average service cost and a lower bound constraint on the average utility, in the asymptotic regime where the average service cost constraint is made arbitrarily close to the above minimum average service cost. The results are presented in J1, C4, and C5. This is also presented in the thesis.
Reliable and delay-optimal communication of bursty sources over communication channels (DRDO-IISc programme on mathematical engineering) An understanding of communication networks is possible only if queueing-theoretic and information-theoretic aspects of communication are considered as a whole. Classical information theory assumes that sources always have information to send and characterizes performance through metrics such as capacity regions and the asymptotic dependence of error probability on coding length. But real world sources produce information that is bursty in nature. The bursty nature of sources has two main implications : a) a bursty source uses the channel resources intermittently thus presenting a time varying channel to the other sources in the network, and b) the random nature of information production and its transmission leads to queueing at the transmitters and resource scheduling problems. In this project, we restricted ourselves to understanding the implication (b) by considering simple discrete time point-to-point channels. We studied the queueing process at the transmitter of a point-to-point channel with a refined information theoretic model, based on the channel coding error exponent, for the transmission scheme. For point-to-point channels using variable length block codes, that guarantee a constant error probability per transmission, we obtain an approximately average-delay-optimal policy, presented in C1 and T1. We also obtain analytical upper and lower bounds on the minimum average delay, presented in C3 and T2. For point-to-point channels using block coding we obtain the exponential decay rate of average error probability with average delay for both fixed and variable length block coding in C2 and T3. Furthermore for fixed block length coding schemes we obtain upper and lower bounds on the tradeoff in C2. In T4 we show that using streaming codes the exponential decay rate of average error probability with average delay can be improved from what that is achieved by block coding schemes. We then consider the problem of allocating transmitter power, for channel inversion, for a slow fading point-to-point channel in T5. Approximations to the optimal allocation are obtained when the average delay is allowed to be large or equivalently when the available average transmitter power is reduced.
|