Aniket Patil
Welcome to my portfolio!
I am a passionate roboticist with a master's in Robotics Engineering from Worcester Polytechnic Institute. I am interested in Computer Vision, Deep Learning and Robotics Software for Mobile Robots.
I worked with the Perception and Autonomous Robotics Group (PeAR) at WPI with Prof. Nitin Sanket. My research involved Structure from Motion and Optical Flow Networks for perception and faster drone navigation. We work with resource constrained tiny robots with on-board computation and sensing and recently submitted a paper which is currently under review.
I am always up for a conversation about technology, cars and photography! Say hi at [apatil2 at wpi dot edu].
Let's connect!
Publications
Aniket Patil, Mandeep Singh, Uday Girish Maradana and Nitin J. Sanket,
"MinNav : Minimalist Navigation Using Optical Flow For Active Tiny Aerial Robots"
(under review IEEE RA-L)
Work Experience
Graduate Researcher,
Perception and Autonomous Robotics Group, WPI, MA
Research Topic: Depth and Post Estimation from Monocular Camera Motion under guidance of Prof Nitin J Sanket
Designed a network to predict Optical Flow and its aleatoric uncertainty for active vision-based navigation on quadrotors
Enhanced the performance using TensorRT and evaluated against RAFT and depth-based networks over the accuracy, inference speeds and other trajectory metrics
Implemented end-to-end depth and pose estimation network based on SFMLearner using monocular camera input (Python, Pytorch)
August 2022 - Present
Robotics Engineer Intern,
DEKA Research and Development Corp., NH, USA
Working with Perception team on trajectory prediction of road agents for FedEx autonomous delivery robot (Roxo)
Analysed the proprietary IRL model performance trained on nuScenes dataset and proposed filters based on analysis of predicted image plots and velocity profiles to eliminate bad predictions
May 2022 - August 2022
Projects
Minimalist Navigation using Optical Flow for Aerial Robots
Research at Perception and Autonomous Robotics Group, WPI under the guidance of Prof Nitin J Sanket (on-going)
Designed a multiscale network for real-time Optical Flow and Uncertainty prediction on an aerial robot with a monocular camera and on-board compute
Demonstrated successful experiments to fly the drone through environments with static and dynamic obstacles, or through gaps
Depth and Pose Estimation from Monocular Camera Motion
Research Project at Perception and Autonomous Robotics Group, WPI under the guidance of Prof Nitin J Sanket
Studying the state-of-the-art implementations on depth, pose and optical flow estimation from the motion of Monocular camera in static scenes
Studying the impact of Uncertainty in the prediction of these parameters and how we can model it to improve predictions
Studied classical computer vision techniques used in visual odometry for Autonomous Vehicles on KITTI dataset
Implemented frontend of stereo odometry pipeline using 3D-2D triangulation of matched features
Implemented RangeNet architecture on spherical projection images of LiDAR Point Cloud (Python, Tensorflow)
Modified the architecture by adding Pixel Shuffling to reduce trainable parameters from 50M to 42M, which reduced the training time by 17mins/epoch on the entire KITTI dataset, with a 5-6% drop in accuracy
Observed better detection of smaller classes such as pedestrians and road signs
Implemented autonomous navigation with dynamic obstacle avoidance using Lattice Planner (global) and Timed Path Follower (local) in hospitals/warehouse environments
Achieved smoother trajectories and better social navigation as compared to the baseline implementation using Base Global Planner and TEB Local Planner
Education
Master of Science,
Robotics Engineering
(2023)
Coursework: Computer Vision, Deep Learning, Motion Planning, Reinforcement Learning, Robot Control, Vision Based Manipulation, Robot Dynamics
Bachelor of Technology,
Instrumentation and Control Engineering (2019)
Coursework: Digital Electronics, Signals and Systems, Digital Control Systems, Digital Signal Processing, Industrial Automation, Micro-controllers