Optimality and performance evaluation of algorithms in communication systems using probabilistic methods

April 15, 2019 – /,

Some key steps in today’s communication systems involve decision making such as scheduling, batching and placement. A common objective among these processes is to achieve optimality or near-optimality, often in the interest of time or resource. We seek to address these optimality problems by suitably modeling the underlying system and the class of applicable algorithms in question. The model should capture the dynamicity of the system and also let us make accurate predictions of the behavior of the system when operated using the relevant algorithms corresponding to a particular process. The key tools are borrowed from the domain of Queueing theory, Markov Decision Process and Optimization. Relevant examples include:

  • How to split chunks over multiple available paths in a network to maximize user experience?
  • How to monitor flows to identify congestion when one has limited monitoring resources?
  • How to batch queries in a closed database system to maximize throughput?

download corresponding tendering

Keywords: Queuing theory, Optimization

Research Area(s):

Tutor: Kar,

Open Theses