Session: 12-11: Software-Defined Vehicles, Digital Twins, and Connected AI Platforms

Paper Number: 197974

197974 - Physics‑informed Digital Twin Learning and Controls for Propulsion Systems in Sdvs 

Abstract: 

Software‑Defined Vehicles (SDVs) are transforming xEV propulsion by enabling continuous software‑driven updates, data‑centric optimization, and adaptive control across the vehicle lifecycle. Achieving this vision requires propulsion models that combine physical interpretability with the flexibility of machine learning. This talk presents a unified framework for physics‑informed digital twin learning, designed to support real‑time estimation and control across a broad range of xEV propulsion architectures.

As an example in this talk, the digital twin models cell‑level electrochemical behavior using a hybrid physics‑guided learning structure. This approach captures nonlinear voltage response, power capability, thermal interactions, and degradation evolution while remaining efficient enough for real‑time propulsion applications. The resulting model provides a robust foundation for energy‑aware control and predictive decision‑making under diverse operating conditions.

A key solution method within this framework is a Physics‑Informed Koopman learning approach, which lifts nonlinear battery dynamics into a constrained linear space. This enables physically consistent estimation of propulsion‑critical energy states such as State of Charge (SOC), while adapting to aging, sparse sensing, and real‑world disturbances. Connectivity‑enabled data further enhances scalability and robustness across fleets of SDVs.

Together, these elements establish a pathway toward self‑updating, intelligent propulsion systems that improve efficiency, reliability, and lifecycle performance across next‑generation xEV platforms.

Presenting Author: Shobhit Gupta General Motors

Presenting Author Biography: Dr. Shobhit Gupta is a Propulsion Controls Researcher in the Energy & Propulsion Systems Research Lab at General Motors (GM). He earned his Bachelor of Technology degree from Indian Institute of Technology Guwahati, India in 2017, the MS and PhD degree in mechanical engineering from The Ohio State University, Columbus, OH, USA, in 2019 and 2022 respectively. His doctoral research focused on perturbed dynamic programming for predictive energy optimization in connected and automated vehicles. Dr. Gupta’s research interest at GM includes data-driven estimation, dynamic optimization, and predictive controls applicable to DC Fast Charging, SOX estimation, energy and thermal management of Connected xEVs. The impact of his research is documented by 30+ publications and 10+ U.S. patents. Dr. Gupta is a member of SAE, ASME, IEEE; honors received include the ASME rising star researcher award in 2021 and ASME Best Paper Finalist in 2022 and 2025. Dr. Gupta serves as the Industry Liaison Chair of IEEE Transportation & Electrification Conference (ITEC) and the ASME Automotive Transportation Technical Committee (2025-26).

Authors:

Shobhit Gupta General Motors