Yash Goel

I am an ML/CV engineer at SafeAD working on 3D object detection and HD maps using transformers. During my masters at the University of Bonn, I worked as a student researcher at StachnissLab and did my master thesis at Bosch Research in Renningen, DE.

Before that, I was a research assistant at Robotics Research Center, IIIT Hyderabad. I did my bachelors in mechanical engineering from IIT Roorkee.

CV Scholar Github Twitter

Publications

Semantically Informed MPC for Context-Aware Robot Exploration

Yash Goel, Narunas Vaskevicius, Luigi Palmieri, Nived Chebrolu, Kai O. Arras, Cyrill Stachniss

IROS 2023, IROS WS 2022

Semantics informed costmap prediction for object goal navigation using MPC.

PDF arXiv Video

PIVO: Probabilistic Inverse Velocity Obstacle for Navigation under Uncertainty

P.S. Naga Jyotish, Yash Goel, A.V.S. Sai Bhargav Kumar, K. Madhava Krishna

ROMAN 2019

Probabilistic version of IVO using chance constraints for taking uncertainty into account.

DOI

IVO: Inverse Velocity Obstacles for Real Time Navigation

P.S. Naga Jyotish, Yash Goel, A.V.S. Sai Bhargav Kumar, K. Madhava Krishna

AIR 2019

Inverse version of Velocity Obstacles with ego-frame formulation shown for real-time and mult-agent setting.

arXiv

Projects

Social robot navigation

Course Project, Uni Bonn

Compared human motion prediction techniques - SocialGAN and maximum entropy method with ROS implementation. Real life implementation on KUKA youbot.

PDF Video

Incremental Poisson Surface Reconstruction

HiWi, StachnissLab, Uni Bonn

Making poisson surface reconstruction incremental by using solution from previous frame as dirichlet boundary condition in the laplace equation.

Bebop Navigation in Real Life

IIIT Hyderabad

Implementing dynamic obstacle avoidance on Bebop drone using collision cone constraint. This was done for both holonomic and non-holonomic settings.

Video

Learning MPC controller

IIIT Hyderabad

Investigated learning MPC controller. Using input of future desired states and the current state, control commands for positional tracking are generated. Simulations were done in ROS using RotorS simulator.

Depth based navigation using RL

Bachelor thesis, IIT Roorkee

Learning control policy using DQN for navigation of a quadcopter in AirSIM using depth images to reach the destination point. We trained the policy for two cases: 2D and 3D action space to compare performance. Published at IEEE AUTEEE.

DOI Video

Formula student electric car

IIT Roorkee

Involved in the design and fabrication of our self designed formula style electric race car as Powertrain Head for 2017 car. Introduced the vehicle dynamics team leading the vehicle dynamics model design, active slip yaw controller, etc.

Video

Hydrogen fuel cell modeling

Internship, TATA Motors Research Center Bangalore

The work involved learning about modelling systems, simulating and designing controller for various sub-systems of a fuel cell vehicle.

PDF