Research

Research focus

Service allocation problems are stochastic optimization problems where a limited number of resources (e.g., servers, bandwidth, transmission power, computing cycles, video segments, routes, dispatch orders) must be allocated to meet the demands of users or tasks in the most effective way. They arise in many large-scale real-world contexts—such as smart cities, urban transportation, wireless communication networks, smart grid, etc. The rapid growth of wireless technologies (e.g., 5G/6G, WiFi-7), multi-access edge computing and artificial intelligence are expected to revolutionize the way we allocate services. The key to service allocation in large-scale networks is addressing the following questions:

a) When and where to run the services?

b) How to allocate the services in a fair and cost-effective manner?

c) How to deal with uncertainties in the network?

d) How to evaluate the quality of experience and guarantee it to the users?

e) How to protect the users from data privacy?

The above issues in large-scale service allocation problems can be addressed by modelling them as Markov decision process (MDP). To tackle the complexity, the large MDP is typically solved by decomposing it into smaller subproblems, where each subproblem is associated with an agent in the network. Determining the timeline and interaction for the agents becomes extremely important. Deep learning techniques allow approximations of the objectives and/or policies of the agents. The observations and decisions of the agents should be of fixed length and lead to lower complexity. Fast multi-agent learning strategies are required. New challenges arise especially when an MDP has constraints (e.g., coupling the actions of the agents). My research deals with tackling such problems in service allocation across three areas namely, energy systems, multi-access edge computing and urban mobility.

Figure: Large-scale network architecture.

Intelligent agents are deployed across the network. They have to learn actions based on the behaviours of the clients and servers.

Clients and servers exhibit dynamic spatio-temporal behaviours.

Mobility, privacy and heterogeneity of clients poses new challenges for service allocation

Past research

Research grants

"Solar Energy Investment and Energy Trading Problem ", PES University Internal funding, PESUIRF/ECE/2020/05 (Start date: 23-09-2020, End date: 24-09-2022)

Research scholars

Veena S works on "Proactive Caching and Computing Services in Vehicular Edge Networks". Expected to complete in 2026.
Keerthi G Reddy works on "Charging, Assignment, Navigation and Repositioning of Electric Taxi Fleet". Expected to complete in mid-2027.
Chinmay Hegde started his research on "AI-based resource allocation algorithms for 6G networks" in Jan 2026.
Anuradha Kannan successfully defended her doctoral thesis on "Modeling and Analysis of Demand-Side Management" in 2023. Dr. Anuradha K works as an Assistant Professor at RNSIT, Bangalore.

Research profile

Google scholar: https://scholar.google.co.in/citations?user=_81IkLIAAAAJ&hl=en

PES University: https://staff.pes.edu/nm1243

Google Sites

Report abuse