VISUALISE (2014-Present)


The retina is an extension of the brain and formed embryonically from neural tissue and connected to the brain by the optic nerve. The retina is the only source of visual information to the brain and a uniquely accessible part of the brain suitable for investigating neural coding. The Visual Modelling Using Ganglion Cells project will create a refined understanding of retinal function in natural visual environments by examining the unique role that non-standard ganglion cells play in dynamic visual processes. Through the use of computational modelling techniques, Visualise will build novel theoretical and hardware models of biological retinal ganglion cell types for dynamical vision applications.

Tools used: ROS, Python, R, Matlab
Robots used: Pioneer3-Dx, Summit-XL

RUBICON (2014)


This project creates a self-learning robotic ecology, called RUBICON (for Robotic UBIquitous COgnitive Network), consisting of a network of sensors, effectors and mobile robot devices. Compared to the current approaches which rely heavily on models of the environment and on human configuration and supervision, a RUBICON ecology will be able to teach itself about its environment and learn to improve the way it carries out different tasks. The ecology will act as a persistent memory and source of intelligence for all its participants and it will exploit the mobility and the better sensing capabilities of the robots to verify and provide the feedback on its own performance. As the nodes of a RUBICON ecology will mutually support one another’s learning, the ecology will identify, commission and fulfill tasks more effectively and efficiently. The project builds on many years of experience across a world-leading consortium. It combines robotics, multi-agent systems, novelty detection, dynamic planning, statistical and computational neuroscience methods, efficient component & data abstraction, robot/WSN middleware and three robotic test-beds. Although this was a three year project, I was involved with it only in the third year. My contribution in this project was to the development of Control Layer programs, which receive goals from the Decision Layer (SOFNN-Self Organising Fuzzy Neural Networks) and sends sensory feedback to Novelty Detection Layer (MDP), to control the PR2 in a home environment and to trigger learning of novel activities. This includes SLAM (Simultaneous Localisation And Mapping) of the environment before the robot is able to navigate efficiently, navigation in the environment using the map, handling of different objects (collection and delivery of different items), and visual perception of different objects and communication to other layers using PEIS (Physically Embedded Intelligent Systems).

Tools used: ROS, Python, PEIS, Matlab
Robots used: PR2

Task Allocation Strategies for Multi-Robot Systems (2010 – 2014)

In many situations such as exploration, service robots in home and healthcare environments and surveillance, where many skills are expected from a robotic system it is beneficial to use a multi-robot system than a single robot with all the required skills. The effectiveness of a multi-robot system is achieved only when the robots coordinate and the tasks are allocated based on the environment and the previous allocation of tasks. This coordination and cooperation among the robots can be achieved using an optimal task allocation algorithm which considers the dynamic changes in the environment and the previous task allocations. Although task allocation is a well-researched problem in different domains including multi-robot systems, it has been traditionally addressed in a time extended manner in which the allocation of all tasks is performed before execution of any task is initialised, to obtain optimal allocations. However the optimality of the allocations is not guaranteed when there are unexpected changes in the environment. The major research focus in this project is to develop a task allocation algorithm which is capable of addressing these dynamic changes in the environment to allocate single and multi-robot tasks without affecting the optimality of allocations. Market-based distributed task allocation algorithms are developed for a heterogeneous multi-robot system handling heterogeneous tasks. My supervisors on this research are Prof. Thomas M. McGinnity and Dr. Sonya A. Coleman.

Tools Used: ROS, Player/Stage, Python
Robots used: Pioneer 3DX
More Info: Available here

Generation Optimization (2009 – 2010)


This project involved development of different Electricity Generation Optimization modules for the Energy Optimization Portal (Eltrix), developed by Kalkitech as a common platform under which different power system optimization applications.The major research challenge was to model the generation optimization problem and implementing it using an optimization tool and integration to the Eltrix. The generation optimization problem was modeled with multiple plants, multiple fuels and thermal constraints of fuels along with other common constraints to get a complete model. The optimization model was implemented using OptimJ (an extension of Java for mathematical programming) and was solved using either CPLEX or MOSEK solvers. The Eltrix won the best product award during the Elecrama-2012.

Tools Used: OptimJ, CPLEX, MOSEK, Java
Skills Learned: Generation Optimization

Phasor Data Concentrator (PDC) (2008-2009)

This project involved development of a PDC for processing, storing and retransmitting heavy data streams from PMUs (Phasor Measurement Unit). The measurements are from the PMUs are synchronized with the GPS and is streamed at a high frequency of 60 fps. These measurements are used to monitor the network over a wide area and there by studying and taking precautionary measures against the fault propagation which will eventually lead to a complete blackout. The product developed in this project, Kalkitech’s PDC is the first of the kind in India and boasts of handling 100s of PMUs (Phasor Measurement Units) at a time. The research challenges included designing of PMU simulator, design, development and integration of the PMU data processing and Storage modules to the PDC data streaming modules. This product won the best Research and Technology Award in the India Power Awards – 2009.

Tools Used: C, MySQL
Skills Learned: IEEE C37.118

Short Term Load Forecasting (STLF) using Artificial Neural network (ANN) (2007-2008)

The research challenge in this project was to develop a feed forward neural network to forecast the electric load for a short period such as a few days to a week. This STLF module plays very important role as one of the major inputs to generation optimization applications. Based on the data obtained from Load Dispatch Centres (LDC), an ANN based STLF module was developed and tested.

Tools Learned/Used: Python, Scipy, Numpy, Matplotlib, WxPython
Skills Learned: STLF, ANN (RBF, FF)

Hydro Thermal Co-ordination (HTC) using Particle Swarm Optimization (PSO) (2007)

This project involved the development of a module to coordinate the generation from multiple generation plants some of which are hydel generation plants and the remaining are thermal generation plants.The HTC module developed had to coordinate generation from these different plants optimally. Particle Swarm Optimization was used for the optimization and different constraints pertaining to both generation plants are also considered for the optimal generation allocations.

Tools Learned/Used: Python, Scipy, Numpy, Matplotlib, WxPython
Skills Learned: PSO, HTC