Img

Autonomous Wheelchair

Guide: Dr. Madhav Krishna RRC, IIIT-Hyderabad

Description: Enabling the wheelchair with automatic localization, navigation in known/ unknown map, object detection and recognition to help patients in hospital. This project contains control, environment recognition, route planning, obstacle avoidance, object recognition and object searching in map. This technology is appropriate for people who have lost mobility due to brain injury or the loss of limbs, but who retain speech.Video Link

Img

Swahana Driverless Car

Guide: Dr. Shanti Swarup (Uurmi Systems), Dr. Madhav Krishna (RRC, IIIT-Hyderabad)

Description: Developing a fully autonomous vehicle for urban surroundings with only low-cost sensors(cameras), which includes reconstruction and segmentation from stereoscopic vision, probabilistic framework for mapping, visual odometry, Kalman fusion for Localization, global/local path planning, robust and computationally efficient lane detection, obstacle tracking and SLAM methods. (Team of 12 | Team Lead | Few Results)

Img

Multi-Agent System for off-road Terrain Mapping

Guide: Dr. Madhav Krishna (RRC, IIIT-Hyderabad)

Funded by: Center for Artificial Intelligence and Robotics(CAIR - Ministry of Defense)

Description: Developing Intelligent system to generate a 3D map of a terrain in real time with multiple multi-terrain robots, system includes 3D pose estimation from cameras, laser scanners, GPS-IMU, and wheel encoders; Localization in current map; and Stochastic mapping. With multi-robot system, mapping process will become extremely fast and fully autonomous. For configuring multiple map data to a single unified probabilistic map we are using g2o optimization framework over octomap 3D representation which will keep correcting robot's relative pose in realtime.






Img

Real time Visual Odometry from Stereo Images

Guide: Dr. Madhav Krishna (RRC, IIIT-Hyderabad)

Funded by: Defence Research & Development Organisation(DRDO)

Description: Implementing real time visual pose estimation modules using reprojection minimization approach and one-point RANSAC approach considering vehicle motion model and estimating road plane surface with linear optimization.

Img

Trajectory generation for non-holonomic vehicle with optimization framework

Guide: Dr. Madhav Krishna (RRC, IIIT-Hyderabad)

Description: Developed a trajectory generation method using convex optimization to avoid obstacle under perception and control uncertainty. Parametric form of high ordered-Bernstein curve was used for trajectory representation and Gaussian function to model noise in control and perception.   


Img

Advanced Drive Assistance System(ADAS)

Guide: Dr. Shanti Swarup (Uurmi Systems)

Description: Developing a low-power FPGA based solution for stereo-camera based ADAS, which includes implementation of accurate stereo reconstruction at very high frame-rate, road surface estimation, obstacle segmentation and tracking, obstacle path prediction in real time for forward collision warning to driver. Implementing slanted plane fitting over SGM disparity output for stereo reconstuction on FPGA.

Img

Human Detection and Tracking in 3D space

Guide: Dr. Madhav Krishna (RRC, IIIT-Hyderabad)

Description: Developed a real time algorithm to detect, track and label multiple pedestrian in current RGB-D view frame (mounted over a moving robot like wheelchair for HRI applications). Algorithm uses Histograms of Oriented Gradient (HOG) in current point-cloud frame and previous detection results as a prior knowledge for detection and tracking. Implemented in C++ using PCL and OpenCV libraries on ROS framework for both CPU and GPU(with CUDA) architecture. (Video Link)

Img

HDR Image construction from LDR Images

Guide: Dr. Jayanti Shivaswami (CVIT, IIIT-Hyderabad)

Description: Developed an algorithm to generate HDR(High Dynamic Range) image from 4 LDR images taken with different exposure time from normal camera. Motion of camera(3 DOF) during exposure is estimated for each image by generating PSF(Point Spread Function) for 10 randomly chosen windows. By applying non-uniform deblurring and fusion we recover High Dynamic Range image.(Team of 2 | Report | Results)

Img

Exploration of Surrounding by Mobile Robot

Guide: Dr. Madhav Krishna (RRC, IIIT-Hyderabad)

Description: A robust algorithm has been developed to explore the surrounding of robot and generate an accurate map which can be used for navigation. The main focus of this project was to find the frontiers in existing map to set new goal position such a way that maximum exploration can be done with minimum navigation cost. ROS platform is used for the navigation and mapping on TurtleBot. (Team of 2)

Img

DenseMeter

Guide: Dr. K R Sharma (IIIT-Hyderabad)

Description: Developed a portable device which can determine density of unknown material using acoustic wave properties in material, in both real and complex domain.

In this project we designed a set up to measure the speed of sound using microcontroller (Atmega16) and sonar sensors

Img

vPlay (Virtual Play Ground)

Guide: Daniel Leithinger and Anirudh Sharma (MIT Media Lab)

Description: Developed a Tangible User Interface (TUI) system to interact with virtual objects in real environment. vPLAY is a platform where user can play with virtual objects, cameras are used to detect surroundings configuration and user’s gestures, and projector is used to display object in real environment. For prototype a laser projector, connected with Arduino controlled two servo actuators(for two degree of freedom) and RGB web- Camera are used. As a programming platform MATLAB environment and controlling platform servo-motors connected to Arduino microcontroller are used.

This project was Developed with MIT Media Lab mentors and then displayed in Research and Development Showcase- 2013 in Hyderabad.




Other Projects

  • RoboCon - 2014 : Designed an autonomous robot which climbs ladders and traverses through uneven terrain by grabbing poles for support.
  • Foreground-Background Estimator for surveillance using Background Mixture Model Link
  • Handwritten Text Binarization using power-law transformation Link
  • Localize logo of a TV channel Link
  • NAYANiX – Linux based Operating System for Blind Link
  • Analyzing Deformable Part Models for different objects Link
  • Control System for Inverted Pendulum Link