Most of my work is on expanding and consolidating distributed network architectures considering the significant security-performance trade-offs noticeable in traditional architectures like CQN, CDN, Dyna, OSG etc. I have investigated centralized and decentralized communication schemes employing peer-to-peer and publisher-subscribe policies to disseminate information within a network. Some of my research projects included the design and development of robust non-parameterized protocols for traditional networks like WSN and Grid systems. I built decentralized protocols for such networks by performing hierarchical clustering within the network and used a genetic approach to randomize the cluster nodes after every generation.
I am currently investigating anomaly detection techniques and mitigation schemes for distributed architectures across complex networks like Grid Systems and Supply Chain networks. I am comprehensively exploring Pareto-based soft computing techniques such as Ant-colony optimization, Fuzzy-based GA algorithms, Cuckoo search etc. to reduce computing cost for identifying risk-impact on complex networks. The applications are numerous ranging from building resilient supply chain networks, risk-aware grid systems and optimizing risk propagation in Bayesian networks.
Modelling and Analysis of k-Bounded Queuing Systems in Wireless Sensor Networks
InWirelessSensorNetworks(WSNs), scheduling and resource provisioning strategies have been heavily scrutinized. It is, therefore, an open problem to model and compare existing resource scheduling algorithms and design formal approaches for safety analyses (invariant analysis, state-space analysis etc). This research study introduces PetriNet analysis as a viable approach to check for significant features like fairness, deadlock avoidance and steady-state reachability in WSN. The scope of this research study is to introduce and demonstrate how Petri Net Formalisms and its variants can be employed to the above-mentioned use cases.
Uncertainty Modelling Uncertainty Modelling Approaches in Risk-averse Supply Chain Systems
One of the arduous tasks in supply chain management is to build robust models against irregular variations. During the proliferation of time-series analyses and machine learning, several modifications were proposed to tackle this challenge such as acceleration of the classical Levenberg-Marquardt algorithm, weight decaying and normalization; which introduced an algorithmic optimization approach to this problem. The central focus of these techniques was to minimize the error and optimized brute force approaches to the absolute minimum. These approaches, however, failed to comply with invariants and uncertainties that induced non-deterministic errors, in time-series models. In this paper, we have introduced some convincing methodologies to handle uncertainty and bound the entropy of such uncertainties by explicitly modelling them. We have outlined Pareto Optimization as a feasible method to perform uncertainty modelling for supply chain networks under some apriori assumptions.
One of the main challenges in grid computing is precision scheduling. This paper introduces a regression-based precision timing in asynchronous distributed grid systems, where we lack information gain on shared resources across a spatial distribution. Through simulation case studies and testing, we have shown that precision scheduling is congruent to discretizing time-variant delays across grid systems. We have observed our self-configured model with continuous message signals and successfully added time-variant delays with information gain on previous timesteps. This is crucial for shared resource management and data-intensive applications as well.
A Comparative Study on Statistical and Neural Approaches for Optimising Supply Chain Management (SCM) Systems
The growing dynamism of the contemporary market is a clear indicator that the strategical execution of business operations will become a crucial factor in the upcoming years. Businesses with a central focus for global operability endure different risks in order to attain a marginally suitable revenue. Several case studies over this new field of research have shredded light into the categorical examination of strategies, risks and managerial procedures involved in such supply chain systems. In this paper, the authors have conducted a naive empirical study of supply chain systems with the use of quasi-experimental methodologies to check for Pareto optimal fronts for largely expanding supply chain networks.
Automated Drip Bio-fertigation Monitoring and Control Systems for Agronomic Practices under Protected Cultivation using Wireless Sensor Networks
This project is currently funded by DSIR-PRISM-TOCIC for a tenure of 3 years.
Effective clustering of reliable sensors in Wireless Sensor Networks (WSN) has always been a challenging problem for the orchestration of smart sensing applications. ML-driven and heuristic approaches have provided ephemeral solutions to clustering problems in WSN. However, these approaches assume the probability distribution of node failures to be uniform. In this paper, we have discussed a reliability-based Bayesian model for enhancing the efficiency of smart grids using statistical machine learning approaches like decision tree algorithm. Decision trees are observed to be robust against variable change and account the dynamic topology of the network. We conducted the simulation of a wireless sensor network using OMNet++ and used the probability model to eliminate unreliable nodes. The results of the simulation case study indicate that carefully considering the criteria for entropy reduction significantly reduces the energy consumption and extends the network lifetime.