Projects

For a full list of my projects, see my GitHub.

• ToolFlowNet: Robotic Manipulation with Tools via Predicting Tool Flow from Point Clouds [Code] [Paper, CoRL ‘22]
  • Imitation learning from point clouds, for robot manipulation of deformable objects. PointNet++ Segmentation backbone imitates ~100 human demonstrations of the task.
  • Developed the entire physical experiments for this work, including setting up Rethink Sawyer robot, extrinsic calibration of Azure Kinect depth sensor, writing human data collection system, collecting multiple sessions worth of human demonstrations, deploying trained network on Rethink Sawyer, and adapting training code to work with real data.
  • Tech stack included PyTorch, Azure Kinect SDK, MoveIt, ZeroMQ and Intera API.
• pyResearchInsights: An open-source tool for scientific text analysis [Code] [Paper, Ecology & Evolution 2021]
  • We developed pyResearchInsights an end-to-end, open-source, automated content analysis tool that Scrapes abstracts, Cleans them up, Analyses temporal frequency of keywords and Visualizes themes of discussions using natural language processing.
  • Collaboration with Vijay Ramesh from the Department of Ecology, Evolution and Environmental Biology, Columbia University, and Anand MO from the Nature Conservation Foundation.
  • Package built using Pandas, BeautifulSoup, gensim and spacy.
• Probabilistic Exploration for Unmanned Aerial Teams [Code] [Paper, IEEE CONNECT 2021]
  • Developed exploration strategies for multi-UAV swarm to conduct reconnaissance of survivors in flooded areas. Advised by Professor Debasish Ghose at the Mobile Robotics Laboratory, Department of Aerospace Engineering, Indian Institute of Science.
  • Simulated the weight-based model on a multi-UAV system comprising of 3DR Iris Quadcopters on ROS & Gazebo, using MAVROS/MAVLINK as a communication protocol.
• Weight-Based Exploration for Unmanned Aerial Vehicles [Code] [Paper] [Paper, Elseiver 2020]
  • Developed a weight-based path-planning model for Unmanned Aerial Vehicles involved in search and rescue. Advised by Professor Debasish Ghose at the Mobile Robotics Laboratory, Department of Aerospace Engineering, Indian Institute of Science.
  • Simulated the probabilistic “weight-based” algorithm on 3DR Iris Quadcopters on ROS & Gazebo, using MAVROS/MAVLINK as a communication protocol.
  • Deployed the model on an off-the-shelf UAV running a MAVROS node onboard a Pixhawk.
• Transiently Informed Robotic Swarms [Code]
  • Part of undergraduate thesis research on developing coordination & exploration strategies for robotic swarms involved in search & rescue, inspired by transiently informed ant swarms. Advised by Professor Vishwesha Guttal and Professor Debasish Ghose.
  • Model designed on ROS using Frontier Exploration, Adaptive Monte Carlo Localization (AMCL), and move_base.
  • Exploration stack is being prototyped on a swarm of Clearpath Huskies, each mounted with a LiDAR to map the environment around it in three dimensions.
• IUCN Red List species threats and stressess database [Code] [Paper, Conservation Letters 2021]
  • Generated a database on species threats and stresses by scraping the IUCN Red List.
  • Collaboration with Uttara Mendiratta and Anand MO from the Nature Conservation Foundation.
  • Utilized Selenium Web Driver to automate the scraping procedure and BeautifulSoup to scrape the HTML contents. Pandas was used to build dataframes and interface with a local .CSV file.
• Fracture-Detection System [Code] [Paper, A2IC 2018] [arXiv]
  • Built a fracture detection system to identify fractures and fatiguing in manufactured components, and predictions were made about the same.
  • The system employed OpenCV through which fractures were identified, after being subjected to different filters.
  • The TensorFlow model, based on ImageNet, retrained on a dataset of normal and fractured gear images, was used as an example to make predictions about wearability and test the system.
• Biomimetic Robotic Fish [Code]
  • Built a predatory robotic fish, as a part of Summer Research Internship at the Theoretical Ecology & Evolution Laboratory.
  • The Robotic Fish will be used to research collective behavior and aggregation patterns in Serpae tetra species of fish.
  • Worked under the guidance of Professor Vishwesha Guttal from Center for Ecological Sciences, Indian Institute of Science.
• Differential Drive Donkey Car [Code]
  • Built a small-scale differential drive autonomous car, powered by a TensorFlow based Autopilot, trained on nearly 10,000 images of a track.
  • The Autopilot was used to drive the vehicle autonomously around the track. Check it out here!
  • Research undertaken at the Center for System Design at the National Institute of Technology Karnataka, jointly advised by Professor Gangadharan KV and Professor Pruthviraj Umesh.
• Gesture Controlled Robotic Arm [Code]
  • Built and designed a 3D printed robotic arm, controlled by human gestures using a Microsoft Kinect 360.
  • Open Natural Interactions Library for Processing was used to detect user’s pose and joint angles.
  • Detected angles were translated to joint movements of the Arm by an Arduino Uno.