Profile Picture

Shubh Gupta

Postdoctoral Scholar | Department of Aeronautics & Astronautics, Stanford University

I am a Postdoctoral Scholar in the Department of Aeronautics & Astronautics at Stanford University. I am working with Professor Grace Gao in the Navigation and Autonomous Vehicles Lab (NAVLab) to develop safe and reliable systems for autonomous navigation. My research focuses on developing navigation systems that can meet stringent safety requirements through the interplay of rich environment modeling, integrity monitoring, and multi-sensor processing. I recently completed my PhD in Electrical Engineering from Stanford University, with my thesis titled "High-Integrity Urban Localization: Bringing Safety in Aviation to Autonomous Driving".

Prior to attending Stanford, I was a graduate student in the Department of Electrical and Computer Engineering at the University of Illinois (U of I) at Urbana-Champaign in the GPS Lab, where I designed robust localization algorithms for GPS pseudorange measurements.

Before my time at U of I, I was at Indian Institute of Technology (IIT) Kanpur where I earned a Bachelors of Technology in Electrical Engineering and a minor in Computer Science and Engineering. During this period, I co-led the student team for participating in the Intelligent Ground Vehicle Competition in 2017. Additionally, I was a member of the student team for participating in ABU Robocon in 2015 and 2016.

My CV is available here.

Projects
Neural Radiance Fields in Navigation: Effective representation of the environment is crucial for designing high-performance and reliable navigation algorithms. We explore environment representations based on Neural Radiance Fields (NeRFs) to aid in localization through camera and GNSS data and to enhance path planning efficiency.
Reinforcement Learning for Risk Assessment: Assessing risk in autonomous driving systems requires efficient identifation of important yet rare failures. Our comprehensive framework integrates driving scenarios, sensor models, autonomous driving policies, and novel rare event simulation techniques, offering a complete risk assessment solution. The code is available here, here, and here.
Deep Neural Networks for GNSS Positioning: In urban settings, GNSS positioning encounters complex non-Gaussian errors from signal reflections and blockages. We have introduced Deep Neural Network (DNN) techniques, based on set-transformers and learned high-dimensional least squares, that effectively blend both traditional and modern methods to address this. Code is available here.
Multi-sensor State and Uncertainty Estimation: Urban multi-sensor localization is complex due to diverse and correlated noise sources. We design several robust algorithms to determine both the state and uncertainty when integrating sensors such as GNSS code/carrier phase, camera, and IMU data. See #1, #2 #3 #4 #5.
Map-aided Error Estimation in Camera Localization: Visual localization is affected by noise factors depending on the immediate surroundings of the navigating vehicle. We develop novel algorithms to establish reliable error bounds in localization by strategically correlating camera images with a 3D map of a city.
Journal Publications
Reliable Urban Vehicle Localization Under Faulty Satellite Navigation Signals.
EURASIP 2023 (submitted)
Shubh Gupta, Grace X. Gao
Urban Localization using Robust Filtering at Multiple Linearization Points.
EURASIP 2023
Shubh Gupta, Adyasha Mohanty, Grace X. Gao
Improving GNSS Positioning using Neural Network-based Corrections.
NAVIGATION 2022
Shubh Gupta*, Ashwin Kanhere*, Akshay Shetty, Grace X. Gao
A Particle-filtering Framework for Integrity Risk of GNSS-Camera Sensor Fusion.
NAVIGATION 2021
Adyasha Mohanty, Shubh Gupta, Grace X. Gao
Data-Driven Protection Levels for Camera and 3D Map-based Safe Urban Localization.
NAVIGATION 2021
Shubh Gupta, Grace X. Gao
Autonomous Vehicle Risk Assessment.
Stanford University 2021 (Technical Report)
Robert Moss, Shubh Gupta, Robert Dyro, Karen Leung, Mykel Kochenderfer, Grace X. Gao, Marco Pavone, Edward Schmerling, Anthony Corso, Regina Madigan, Matei Stroila, Tim Gibson
Conference Publications
Neural City Maps for GNSS NLOS Prediction
ION GNSS+ 2023 (accepted)
Daniel Neamati, Shubh Gupta, Mira Partha, and Grace X. Gao
Neural City Maps: A Case for 3D Urban Environment Representations Based on Radiance Fields.
ION GNSS+ 2023 (accepted)
Mira Partha, Shubh Gupta, and Grace X. Gao
Neural Radiance Maps for Extraterrestrial Navigation and Path Planning.
ION GNSS+ 2023 (accepted)
Adam Dai, Shubh Gupta, and Grace X. Gao
Getting The Best of Particle and Kalman Filters: GNSS Sensor Fusion using Rao-Blackwellized Particle Filter.
ION GNSS+ 2022
Shubh Gupta, Adyasha Mohanty, Grace X. Gao
Designing Deep Neural Networks for Sequential GNSS Positioning.
ION GNSS+ 2022
Shubh Gupta, Ashwin Kanhere, Akshay Shetty, Grace X. Gao
Improving GNSS Positioning using Neural Network-based Corrections.
ION GNSS+ 2021
Shubh Gupta*, Ashwin Kanhere*, Akshay Shetty, Grace X. Gao
A Particle Filtering Framework for Integrity Risk of GNSS-Camera Sensor Fusion.
ION GNSS+ 2020
Adyasha Mohanty, Shubh Gupta, Grace X. Gao
Data-Driven Protection Levels for Camera and 3D Map-based Safe Urban Localization.
ION GNSS+ 2020
Shubh Gupta, Grace X. Gao
On the Role of Receptive Field in Unsupervised Sim-to-Real Image Translation.
Machine Learning for Autonomous Driving Workshop at NeurIPS 2019
Nikita Jaipuria, Shubh Gupta, Praveen Narayanan, Vidya Murali
Particle RAIM for Integrity Monitoring.
ION GNSS+ 2019
Shubh Gupta, Grace X. Gao
Saliency Prediction for Mobile User Interfaces.
WACV 2018
Prakhar Gupta, Shubh Gupta, Ajaykrishanan Jayagopal, Sourav Pal, Ritwik Sinha
A Deep Reinforcement Learning Approach to Rare Event Estimation.
arXiv
Anthony Corso, Kyu-Young Kim, Shubh Gupta, Grace X. Gao, Mykel Kochenderfer

Also on Google Scholar
Teaching
In 2020 I was the primary TA for the GPS class AA272C: GLOBAL POSITIONING SYSTEMS and also helped design and teach the first autonomous navigation class AA275: NAVIGATION AUTONOMOUS SYSTEMS at Stanford in 2020 and 2021. Both the classes are steadily growing from 12 students in 2020 to being at full capacity in 2023.