publications

2025

1. SIGEnergy Demo

   Open In-Context Energy Management Platform

   Yikai Lu, Tinko Sebastian Bartels, Ruixiang Wu, and 5 more authors

   In Proceedings of the 16th ACM International Conference on Future and Sustainable Energy Systems, Jun 2025

   Abs DOI

   Recent advances in Foundation Models allow leveraging rich, multi-modal context to enhance performance in a variety of tasks. The management of modern energy systems including renewable energy generation and batteries is a control problem that is inherently context-sensitive due to the complex interplay of future energy demand, electricity prices, energy generation from wind, solar, and other sources. This presents new challenges in model training with existing narrow data sets, and context design. In this contribution we present the Open In-Context Energy Management Platform (OpenCEM), which will provide a rich, open data set of energy system management time series, event information, user-generated context, and a simulator through a public API to evaluate the performance of control algorithms and to provide quality feedback for user-supplied context. The data is obtained from an on-campus PV installation with batteries.

2. IEEECDC

   InstructMPC: A Human-LLM-in-the-Loop Framework for Context-Aware Control

   Ruixiang Wu, Jiahao Ai, and Tongxin Li

   In 2025 IEEE 64th Conference on Decision and Control (CDC), Dec 2025

   Abs DOI

   Model Predictive Control (MPC) is a powerful control strategy widely utilized in domains like energy management, building control, and autonomous systems. However, its effectiveness in real-world settings is challenged by the need to incorporate context-specific predictions and expert instructions, which traditional MPC often neglects. We propose INSTRUCTMPC, a novel framework that addresses this gap by integrating real-time human instructions through a Large Language Model (LLM) to produce context-aware predictions for MPC. Our method employs a Language-to-Distribution (L2D) module to translate contextual information into predictive disturbance trajectories, which are then incorporated into the MPC optimization. Unlike existing context-aware and language-based MPC models, INSTRUCTMPC enables dynamic human-LLM interaction and fine-tunes the L2D module in a closed loop with theoretical performance guarantees, achieving a regret bound of O\sqrt T\log  T for linear dynamics when optimized via advanced fine-tuning methods such as Direct Preference Optimization (DPO) using a tailored loss function.

2024

1. arXivPreprint

   Behavior-Aware Efficient Detection of Malicious EVs in V2G Systems

   Ruixiang Wu, Xudong Wang, and Tongxin Li

   Dec 2024