Venkata Amith Palacherla

Ñ Brooklyn, NY +1-848-***-**** # advzs8@r.postjobfree.com ï in/venkata-palacherla § github.com/venkataPalacherla Experience

Control/Robotics Research Laboratory January 2022 – Present Graduate Research Assistant Brooklyn, NY

• Developed Real-time 3D Semantic SLAM algorithm to map using 2D stereo data. ROS middleware was used.

• Engineered a particle filtering-based active localization system achieving a high localization accuracy of 4 cm.

• Reduced semantic miss-classifications by 25% by creating a fusion-based voxelization filter for Point Clouds.

• Integrated Semantic information into Visual Odometry pipeline to improve performance by 50%.

• Designed and implemented motion planning algorithms, including a solution for providing voice-guided navigation for blind individuals.

Shiv Nadar University August 2020 – July 2021

Research and Teaching Assistant Delhi, India

• Led a team of 10 in developing the ROS-based remotely controlled agriculture robot. Received funding of $5,000.

• Implemented robotics software pipeline to control electric motor, Nvidia Jetson, and Gmapping SLAM.

• Improved mobile robot localization by fusing sensor data from Visual Odometry, IMU, GPS, and wheel. Technical Skills

Programming Languages: Python, C++,C,CMake, Matlab,Arduino Tools/Frameworks: Robotic Operating System(ROS), OpenCV, PCL, PyTorch, TensorFlow, Gazebo, Linux, Eigen Skills: Computer Vision, Deep Learning, Kalman Filter, 3D Geometry, Feature detection, Motion Estimation. Hardware: Embedded Linux Platform, Stereo Camera, Monocular Camera, IMU, GPS, Lidar. Research

• ”Real-Time Visually Assisted System for Indoor Localization and Navigation,” International Conference on Intelligent Robots and Systems (IROS) 2023, Under Review.

• ”Efficient Real-Time Localization in Prior Indoor Maps Using Semantic SLAM,” 9th International Conference on Automation, Robotics and Applications (ICARA), February’23. [Link] Projects

Attention based Visual Odometry Deep Learning, Python, PyTorch [Link] Fall’22

• Collaborated with team of 2 to develop temporal attention based visual odometry.

• Achieved translation & rotation error of 0.05m & 2 . Created a dataset to test VO in feature sparse environment. Optimizing ResNet for Maximum Accuracy on CIFAR-10 Deep Learning, Python,PyTorch [Link] Fall’22

• Collaborated with team of 2 to develop a neural network with ResNet backbone with less than 5 million parameters.

• Achieved accuracy of 95.55% on CIFAR-10 dataset.

• Analyzed impact of number of layers and channels in ResNet on model performance by training 15 models. Acrobatic Control of a 2D Quadrotor with LQR & iLQR Optimal Control, Python [Link] Fall’22

• Modelled dynamics of multirotor drone and determined the stable points of the system.

• Designed LQR controller for the drone to follow a predefined path with 0.01m accuracy in disturbances.

• Implemented iLQR controller for the drone to perform acrobatic manoeuvres, successfully executed 180 and 90 flips. Exploring Q-Learning for Pendulum Swing-Up Control Reinforcement Learning, Python [Link] Fall’22

• Developed Q-learning based controller for the multivariate system.

• Examined the effectiveness of the controller by learning a torque control policy for swinging up an inverted pendulum.

• Improved learning by adding ϵ-greedy policy and analyzed the learnt policy for different control limits and learning rates. Visual Pose Estimation of Drone Kalman Filter, Sensor Fusion, Matlab, Optical Flow [Link] Spring’22

• Estimated twist of drone with accuracy of 85% by detecting keypoints and KLT tracking between camera frames.

• Fused Optical Flow and IMU data using Extended Kalman filter to calculate the drone’s pose with error of 0.05m.

• Implemented Unscented Kalman Filter for state estimation of drone using non-linear process and update model. Education

New York University May 2023

Master of Science in Mechatronics and Robotics New York, NY Shiv Nadar University May 2020

Bachelor of Engineering in Mechanical Engineering Delhi, India

Contact this candidate