Varun Nayak

As part of a course project (AA274: Principles of Robot Autonomy) at Stanford, we developed ROS packages for a turtlebot to autonomously explore and navigate through a mock environment. It would "pickup" and "deliver" food items that it saw along the way.
For autonomous exploration, we used a costmap overlay over a map generated by EKF-SLAM package gmapping and a modified Djikstra algorithm that found the closest unexplored part of the map.

Our team designed and developed an autonomous differential drive robot from absolute scratch in the ME218B course project. The robot had to pickup, sort and dispose off coloured balls in specific zones while navigating an arena with various provisions for localization such as beacons and lines.
Our robot relied heavily on navigation using beacon detection and inductive field sensing. Decision making was carried out by implementing state charts for various tasks and imparting… Show more

Our team designed and developed an autonomous differential drive robot from absolute scratch in the ME218B course project. The robot had to pickup, sort and dispose off coloured balls in specific zones while navigating an arena with various provisions for localization such as beacons and lines.
Our robot relied heavily on navigation using beacon detection and inductive field sensing. Decision making was carried out by implementing state charts for various tasks and imparting intelligence. 

Show less

The project was an arcade-style interactive game that contained several user-inputs (both digital and analog) and audio-visual feedback to keep the user engaged. The Texas Instruments 'TIVA' was used as the microcontroller for the project. An events and services framework was used to implement the desired behaviour of the interactive game.

-Worked on vision-based autonomous landing of AR.Drone.
-Implemented and thoroughly tested PID with gain scheduling for high-level control using only a single monocular camera and grey-scale median filtering.
-The UAV was able to land on stationary, inclined and moving platforms with an accuracy of a few centimetres.

- Conceptualized, designed and tested a novel combination of a noise-based Gaussian filter and a sigmoid function-based sensor fusion method to estimate proximity data from low-cost sensors.
- The filter output was used to drive a PID controller for efficient target tracking of moving objects by a quadrotor.
- The experimental results demonstrated tracking of an object moving with a speed of 1.5 m/s with an error of less than 15 cm.
- Paper submitted to an international conference… Show more

- Conceptualized, designed and tested a novel combination of a noise-based Gaussian filter and a sigmoid function-based sensor fusion method to estimate proximity data from low-cost sensors.
- The filter output was used to drive a PID controller for efficient target tracking of moving objects by a quadrotor.
- The experimental results demonstrated tracking of an object moving with a speed of 1.5 m/s with an error of less than 15 cm.
- Paper submitted to an international conference organized by the International Federation for Automatic Control (IFAC). Show less

-Was hired as a intern for a project which aimed at optimising HPCL's throughput of white oil products using only existing facilities and demand constraints.
-This solution was presented to Chief Engineers at HPCL and was successfully implemented thereafter, saving a margin of $70,000 when compared with an existing solution.