The Goal we are chasing.

Autonomous Flying Vehicles

We aim to move the state-of-the-art in aerial robotics forward. Our drones, when ready, would be capable of autonomous flight, localization in GPS denied environments, tracking and interacting with mobile ground robots as well as other drones, and much more!

International Aerial Robotics Competition

IARC is the longest running collegiate aerial robotics challenge in the world. Entering its third decade, the competition continues to tackle challenges that are currently impossible for any flying robots owned by government or industry. We aim to participate in its next iteration.

Project Description

Details of the work done.

ARK aims to develop a flexible aerial robotics framework which can be easily used to control aerial vehicles. Eventually, we want to move to multiple decentralised aerial robot swarms in an outdoors setting. Currently, we have designed an architecture for IARC which we plan to optimise iteratively for participation.

The software team is responsible for developing algorithms in computer vision including localisation, object detection, tracking, path planning and 3D obstacle avoidance along with artifical intelligence algorithms for exploration and herding. Current Areas of focus are:

Object Detection and Tracking

Object detection and tracking is widely used with applications in monitoring and surveillance. In IARC, the task requires us to guide ground robots towards a line. We plan to use a fisheye lens to see the whole arena and then track the ground robots, maintaining the trackers even when they are out of field of view. We are using Kalman filter for the tracking on the linear model of motion of the ground robots.

Ground Truth Validation

A true value of the robots position and orientation is a necessary metric to compare our algorithms of self estimation, detection and tracking. We are using two high-definition cameras fixed on the arena frame to get the depth of the desired object which will then be transformed into a global coordinate system of the arena.

Grid Based Localization

IARC is held in an indoor GPS denied environment so localization is a major challenge. IARC provides grids on the ground which can be detected and used to localize.

Optical Flow Localisation

Optical flow is the pattern of apparent motion of objects, surfaces, and edges in a visual scene caused by the relative motion between an observer (an eye or a camera) and the scene. From the optical flow of a video feed, a path and trajectory of robot motion can be inversely computed, which is one of the most widely used ways of localisation in GPS denied environment.


IARC has an interesting AI aspect in it's problem statement which is to herd robots past the green line and avoid them crossing the red line. This brings an additional constraint of targetting the robots moving towards red line first and optimizing the attack behavior so that maximum robots can be herded across the line.

Simulation on Gazebo

Aerial Robots are inherently difficult to control and testing on real robots is not always feasible. Hence, a simulation platform on Gazebo which uses the Quad model and physics and also simulates sensor noise and environment is an important step.

The Controls Team aims towards a smooth motion of the aerial robot in space and that they attain the desired speed and orientation avoiding obstacles. Current research includes:

Two Layered PID

An aerial robot can control it's position in the form of roll. pitch and yaw but in the world frame, It gets an input or feed back in the form of X,Y and Z. So, a cascaded PID with self adjusting gains and parameters is the direction towards which we are working. We have got a naive PID over RPY and XYZ working on AR Drone and are working to shift to self-adjusting parameters.

Low Level Controller

Low Level Controller includes the embedded system on the robot end where commands from the high level controller and processed and converted into Motor Commands. We have developed a MATLAB Simulink model for our own chassis and working on that to develop on of our controller. Also, we are trying to get Telekyb Near Hover Controller from Max Planck Institute of Biological Cybernetics to run on our own boards which control the motors.

Self Estimation

A robot should know where is it. Knowing the location of the robot in the world using onboard sensors is a major challenge in robotics. We are using IMU data along with vision information and SONAR sensors to fuse the information and get an estimate of our own position. We are constantly trying to improve this by comparing it with our true value setup and move the state of the art ahead.

Path Planner

The robots in order to perform some tasks needs to move from one place to another. During it's course of motion, it may deviate and the information from estimator and vision is noisy. Moroever, in order to fulfill the IARC problem statement, we need a time dependent trajectory generated and followed to optimally reach a ground robot and perform any action on it.

Media Coverage

Recent media articles about us.

Times of India Economic Times Indian Express Business Standard
Daijiworld Naya India


Relive the History.

Our Amazing Team

The Force to reckon with.

Prof. Somesh Kumar

Principal Investigator, Department of Mathematics

Prof. Jayanta Mukhopadhyay

Mentor,Department of Computer Science & Engineering

Prof. Dilip Kumar Pratihar

Mentor, Department of Mechanical Engineering

Aditya Agarwal

Founder, Alumnus

Chaitanya B

Founder, Alumnus

Avinash Ruchandani

Alumnus, Software Team

Githin John

Founder, Alumnus

Mrinal Mohit

Founder, Alumnus

Soumyadeep Mukherjee

Founder, Alumnus

Jit Ray Chowdhury

Mentor, Alumnus

Prasann Jain

Alumnus, Controls Team

Keshav Sarraf

Alumnus, Controls Team

Sairam K

Alumnus, Software Team

Rishal Raj

Alumnus, Controls Team

Kumar Krishna Agrawal

Alumnus, Software Team

Vishnu Sharma

Alumnus, Controls Team

Gaurav Gardi

Alumnus, Controls Team

Manash Pratim Das

Head, Software Team

Sourish Ghosh

Member, Software Team

Ashwary Anand

Member, Software Team

Aman Chandra

Member, Controls Team

Amit Kumar Pathak

Member, Controls Team

Krishnakant Deshmukh

Member, Controls Team

Shivang Agrawal

Member, Controls Team

Vivek Mudgal

Member, Controls Team

Gaurav Suryawanshi

Member, Software Team

Akshay Jain

Member, Hardware Team

Akshat Pandya

Member, Software Team

Manad Mishra

Member, Software Team

Aditi Singh

Member, Controls Team

Praneet jain

Member, Software

Aman Modi

Member, Hardware and Embedded

Anvee Naik

Member, Software Team

Shubhika Garg

Member, Software Team

Vidit Goel

Member, Software Team

Palak Harwani

Member, Controls Team

Aryan Jaiswal

Member, Software Team

Biswajit Ghosh

Member, Software Team

Parakh Agarwal

Member, Software Team

Abhinav Ukey

Member, Hardware and Embedded Team

Contact Us

We are always open to new members. Fill this and we will have a small task to see if we are good enough for you!

(Please Add Your CV Link/Website in this)