I am a PhD student at the Department of Mechanical Engineering - Texas A&M University. I am working under the guidance of Dr. Srikanth Saripalli I completed my Btech and Masters in Computer science majoring in Robotics from International Institute of Information Technology Hyderabad under the guidance of Dr. K. Madhava Krishna. I worked as a full time research engineer at Uurmi Systems on Autonomous Driving specifically I was leading the teams working on Localization(5 members) and Navigation (3 members) subsytems.

Contact

Akhil Kumar Nagariya,
4302, college main street,
Apt 304, Bryan
Texas, 77801

Research Interests

My primary research Interests include Mobile Robotics, Robotic Vision, Machine Learning, Optimal Contorl and Autonomous systems. My MS thesis is on Human Intention prediction and its incorporation into the Navigation system of a mobile robot for its safe navigtaion. Click here to see my CV which summarizes my skills and work.

  • 2015

    Akhil Nagariya, Bharath Gopalakirshna, Arun Kumar Singh, Krishnam Gupta, K. Madhava Krishna, "Mobile Robot Navigation Amidst Humans with Intents and Uncertainties: A Time Scaled Collision Cone Approach." ,54th IEEEConference on Decision and Control 2015 [Pdf]
  • 2016

    Siddharth Tourani, Sudhanshu Mittal, Akhil Nagariya, Visesh Chari, Madhava Krishna,"Rolling Shutter and Motion Blur Removal for Depth Cameras". ICRA 2015 [Pdf]
  • 2012

    Karthik Desingh,Akhil Nagariya, K Madhava Krishna,"Viewpoint based Mobile Robotic Exploration aiding Object Search in Indoor Environment". ICVGIP 2012 [Pdf]
  • Human motion modelling and Navigation in indoor environment

    Navigating among humans requires a robot to understand human intentions.In this project, we developed an HMM-based approach to model human intentions and proposed a probabilistic variant of time scaled collision cone which allowed the association of space of collision free velocities with a confidence measure. We also developed an optimization framework which extracts a specific solution from the entire solution space that achieves an elegance balance between the objective of minimizing risk and ease of avoidance maneuver. I presented this work at CDC 2015. Refer to the paper [Pdf] for more details.

    Autonomous Navigation of Wheelchair

    We developed a wheelchair capable of SLAM, localization, and autonomous navigation in indoor environments. The localization algorithm uses AR tags to relocalize the wheelchair in the case of uncertainty explosion and lost pose. I also developed a local navigation algorithm which uses Bernstein's polynomials to generate kinematically consistent trajectories. Finally, the trajectory minimizing a cost function designed to provide the tradeoff between comfort and collision is selected from this set.

    Mobile Robot Exploration and Object search

    This work was done in colaboration with Karthik Desingh. It was accepted in ICVGIP'12. First the map of the environment is provided to the robot and then it is asked to find an object in the environment. The robot plans it's path so that it maximizes the probability of sighting an object.The video shows the demo of turtlebot exploring the environment to find an object the video also shows the object detection module that was built using PCL only depth information is used in object detection. More detailed description of the approach can be found in the paper click here or the image to see the video

    Global Planning in Open street Map

    As a Research Engineer at Uurmi Systems, ,I Implemented a global path planning algorithm for autonomous driving. Given the discrete set of waypoints to follow, it generated a plan consistent with the open street map. I used A* for path planning and Kd-tree for optimizing searches for finding nodes in the graph that are closest to a particular waypoint. It also generates different center Lanes that the vechicle can follow based on the information provided by open street map. The code is written in Python and has been integrated successfully with the Behavior Layer of the vechicle.

    Robosoccer

    Developed six wireless soccer playing robots for the event Robosoccer organized in Felicity 2013. The robots were based on defferential drive and could be contorlled by the joysticks all the robots have a kicker installed which was used to pass or shoot the ball unfortunatley because of some communication problem the event was cancelled click here or the image to see the video

    Six Legged Robot using two motors

    This is an on going project we are still in it's development phase. The Robot uses four, four bar mechanism two each side lying on a plane. both sides are controlled independently by a motor depending on the rotation speed of the two motors it can turn left or right further we we will try to come up with a stable mechanism so that it can carry objects from one place to other

    Montecarlo Localization

    Implemented monte carlo localization in simulation on matlab the robot simulated was differential drive which is moved around in the 3d environment we simulated 24 lasers to get the mesurments the image shows how the localization works in this implemetation to sample the pose we followed velocity motion model as described in probablistic robotics book

    Online Quiz Application

    Developed an online quiz application as my project work under the course Software Systems and Application Development. This was a team project (5 members). We used HTML, CSS, Java script and PHP - MYSQL. Moderator can add different questions and their difficulty level. As the question is solved by more people its difficulty level decreases and the students will get less marks. Students can also see their Ranks amongst each other

    3D Town Model

    This project is developed under the course Graphics we developed a 3D town model and different means of transport to travel using c++ opengl library.We also implemented lighting, Day light and it's direction depends on the time of the day.Matarials have different appearance on differentkind of lights many cameras are implemented to see the model from different-different location like inside the car or the halicopeter view from overhead.

    EKF Localization in Amigobot

    Implemented EKF localization on Amigobot using Aira to control and connect to Amigobot , Real Map was generated manually and corresponding to this map the location of landmarks and the measurment correspondence was give to the robot the code is implemented in c++ which calls matlab for the localization module and executes the control command obtaind from matlab

    RRT for RPR robotic arm

    Implemented EKF localization on Amigobot using Aira to control and connect to Amigobot , Real Map was generated manually and corresponding to this map the location of landmarks and the measurment correspondence was give to the robot the code is implemented in c++ which calls matlab for the localization module and executes the control command obtaind from matlab

    Address:

    Akhil Kumar Nagariya,
    4302, College main street,
    Apt 304, Bryan,
    Texas, 77801

    Connect Using:

    Email: akhil.nagariya@gmail.com
    Mob: (+1) (979) 422 2630

    Social Media: