Distributed Estimation of Largescale Systems:
Networked estimation is where a
network of agents is tasked to estimate the state of a dynamical
system by sharing information over a sparse communication network;
Recently, consensusbased estimation has been proposed as a distributed solution.The agents implement a messagepassing algorithm on their measurements between every two successive timesteps of the system dynamics. However, when the network is unable to consensus due to, e.g., resourceconstraints or fast system dynamics, distributed solutions with finite information exchanges have been proposed, at the price of sharing statepredictions. In this scenario, one first has to guarantee the existence of a communication network that will result into a bounded estimation error, i.e., networked observability.
As the main contribution we, first, provide a novel construction
of the distributed estimator and distributed observability from the
first principles. Secondly, we describe a graphtheoretic agent
classification that establishes the importance and role of each
agent towards estimation. In particular, we classify agents as
Type\alpha as the most crucial agent, Type\beta as the less
crucial agent, and Type\gamma as the noncrucial agent. Thirdly,
to achieve distributed observability, we derive the necessary
and sufficient conditions on the agents' communication network based on the
aforementioned agent classification. We prove that for distributed
observability every Type\alpha agent has a direct connection to
every other agent, and every agent has a directed path to every
Type\beta agent. The combination of these two requirements defines
the multiagent network. Advances in the current stateoftheart: 1. Only one information exchange is allowed among the agents. 2. The results are structurebased in contrast to widely used algebraic (rankbased) tests. 

3. When the system matrix is structuredrank deficient, then no
agent communication network can guarantee observability with
agreement in the statepredictor space alone, and hence, fusion in
the observation space is required.
Distributed Randomized Shortest Path Problem:
Advances in the current stateoftheart: 1. The algorithm is distributed in the sense that each node only need to know its immediate neighbors. 2. The graph topology is unknwon but get discovered gradually for the source node. This rises application in Network Tomography. 

Finitetime Consensus and Discrete Resource Allocation: (M.Sc. Thesis)
In this research, we explore the required conditions for reaching consensus in finitetime. We derive the necessary condition as having nonLipschitz consensus function at the origin, e.g. sign function. Since applying such condition is not feasible for realworld actuators, smooth approximations such as tanh() function are proposed to acheive a consensus boundary in finitetime. Lyapunov stability analysis is performed to prove the convergence. One application of this consensus protocol is in discrete coverage control over a convex polygon, also regarded as discrete source allocation problem.
A group of mobile agents deploying over a convex area to optimally allocate resources: the blue circles represent two centers of resource density distribution. Each agent (shown as black square) is assigned with resources (yellow dots) in its Voronoi cell. In the figure this assignment is represented by a yellow line connecting the resource to the agent. 
Kinematic and Dynamic Analysis of Mechanisms:
As part of my M.Sc. and B.Sc. research, I was working on projects in the field of robotics and mechanism design, some of them presented below:
Parallel robots:
Parallel mechanisms are known for high precision, fast tracking
and high load to weight ratio. In this course project a 3pod
Delta robot is cinematically analyzed (inverse kinematics) for
fast machining process. Specifically, the tradeoff between speed
and geometric workspace is examined and compared to the
industrial series manipulators. Steering Mechanism (B.Sc. Thesis): When a car is turning over a curve, the inner wheel turns sharper than the outer one. This is because the entire car must turn around a common center point. The purpose of the steering mechanism is to point the wheels to follow such directions. In this project, 4Linkage steering mechanism is considered. The error for different rotation angles is obtained for different size of linkages and the result shows a consistency with Ackermann steering geometry. 
