Open Positions



Hardware design and follow a known aggressive trajectory on a quadrotor with on-board sensing

Description: Quadrotors are very agile vehicles and their agility shines through when executing high speed trajectories. This has been done using motion capture system which gives “perfect” estimates of the position and orientation of the quadrotor. Executing these high speed trajectories using on-board sensing is impressive and very useful for various tasks like drone racing, obstacle avoidance and flying through narrow gaps.
Goal: Design a quadrotor platform with onboard sensors and implement aggressive trajectory following using only a monocular camera and an IMU with all processing done on-board.

Contact: Nitin Sanket (nitinsan[at]terpmail[dot]umd[dot]edu) and Chahat Deep Singh (chahat[at]terpmail[dot]umd[dot]edu)
Skills Required: Linux, C/C++, Python and ROS.
Thesis Type: Semester Project/Independent Study
Job Opportunities: Upon successful completion of the project, students will have a good chance of getting a job at DJI and Parrot.



Benchmarking Neural Network Compression/Speedup Methods

Description: Deep Neural Networks (DNNs) have advanced the state of the art in many computer vision algorithms in the recent times, however they tend to be both memory and computationally expensive prohibiting them from usage on small mobile computing platforms.
Goal: Benchmark different major neural network compression and speedup methods presented in the last two years.

Contact: Nitin Sanket (nitinsan[at]terpmail[dot]umd[dot]edu) and Chahat Deep Singh (chahat[at]terpmail[dot]umd[dot]edu)
Skills Required: Linux, Python, TensorFlow, PyTorch.
Thesis Type: Semester Project/Independent Study
Job Opportunities: Upon successful completion of the project, students will have a good chance of getting a job at computer vision and deep learning companies.



Design and implementation of gripper for aerial grasping

Description: Grasping from a Quadrotor is a very challenging and an understudied problem. This entails working with an unstable platform for grasping and controlling the system taking into account the dynamic changes.
Goal: Designing a gripper on both software and hardware platform for efficient grasping of a table-top object to be mounted on a quadrotor.

Contact: Chahat Deep Singh (chahat[at]terpmail.umd.edu) and Nitin Sanket (nitinsan[at]terpmail.umd.edu).
Skills Required: Linux, CAD, Python, experience with ROS, C/C++ is a plus.
Thesis Type: Semester Project/Independent Study
Job Opportunities: Upon successful completion of the project, students will have a good chance of getting a job at Amazon and other drone delivery companies.


Deep Learning based Robust Grasping of objects using UR-10, Sawyer and Baxter

Description: Grasping is a fundamental problem which need to be solved by humanoid robots or robotic arms. Grasping traditionally involves end effector planning coupled with perception and complex controls.
Goal: Using Deep Learning based models to perform robust grasps. You’ll be using the DexNet 2.0 from UC Berkeley to implement these robot grasps.

Contact: Kanishka Ganguly (kganguly[at]terpmail[dot]umd[dot]edu))
Skills Required: Deep Learning, Linux, Python and ROS.
Thesis Type: Semester Project/Independent Study/Masters Thesis
Job Opportunities: Upon successful completion of the project, students will have a good chance of getting a job at Amazon and any other robotic arm companies or robotics automation companies.


Operating Home Appliances using the Baxter Robot

Description: Grasping is a fundamental problem which need to be solved by humanoid robots or robotic arms. Grasping traditionally involves end effector planning coupled with perception and complex controls. In this project, you will be working on operating refrigerator and microwave oven in particular. Check [1] and [2] videos for reference. The codebase is available and you will be responsible for improving the robustness of the system.
Goal: Operating the refrigerator and microwave oven using Baxter.

Contact: Cornelia Fermuller (fer[at]cfar[dot]umd[dot]edu))
Skills Required: Proficiency in ROS, C++, Linux and Python.
Thesis Type: Semester Project/Independent Study
Job Opportunities: Upon successful completion of the project, students will have a good chance of getting a job at Amazon and any other robotic arm companies or robotics automation companies.


           

 info[at]prg.cs.umd.edu