Advanced Mobility & Augmented Intelligence Lab

News

[Fall 2026]
  • One paper led by Wei Hu has been accepted by EMNLP’26 Main Conference. The acceptance rate is 2,719/17,669=15.4%. Congratulations!
  • One paper has been accepted by ICDM’26. The acceptance rate is 289/1,450=19.93%. Congratulations!
[Spring 2026]
  • Invited to serve on 2027 ACM e-Energy‘s TPC.
  • Adele Chinda will join Lenovo in Chicago as AI Research Intern this summer. Congratulations!
  • One paper led by Hongyu Ke has been accepted by ICML’26. The acceptance rate is 6,352/23,918=26.6%. Congratulations!
  • Excited to kick-off the research collaboration with Heartfelt Souls. This research partnership advances the science behind memory preservation, voice modeling, and emotionally intelligent AI, pushing the boundaries of what Soul LM can understand and express.
[Fall 2025]
  • I am honored to be awarded a Distinguished Member of the IEEE INFOCOM 2026 TPC!
  • Received $50,000 research grant from InfoTech Labs, Toyoto Motor North America. Thanks Toyota!
  • I am invited to serve on TPC of IEEE ICDCS 2026 @ Seoul, South Korea.
  • Our paper lm-Meter: Unveiling Runtime Inference Latency for On-Device Language Models, has been named a Best Paper Award Finalist (Top 5%) at ACM/IEEE SEC 2025!
  • Xiaolong was selected as one of the three recipients of the 2025 Martin D. Fraser Graduate Student Conference Travel Award, which is a merit-based accolade that supports CS PhD students at GSU with conference travel. 
  • Two papers in our group were accepted by ACM/IEEE Symposium on Edge Computing (SEC’25).
  • We’ve kicked off our research collaboration with U.S. Army Research Laboratory (ARL) on hardware-aware NAS for mission-critical edge applications.
  • I’m opening a new course: CSc 4980/6980 Efficient AI, this FallWe’ll introduce cutting-edge techniques in model compression and hardware-aware design to make AI inference and training faster, more efficient, and scalable, with applications to emerging models such as LLMs and Vision Transformers.
  • I am invited to serve on TPC of IEEE INFOCOM 2026 @Tokyo, Japan.
  • Welcome on-board! My 4th Ph.D. student Wei Hu joined AMAI Lab and will be working on energy-efficient on-device LLMs.
[Spring 2025]
[Fall 2024]
[Spring 2024]
[Fall 2023]
  • Undergraduate researchers from AMAI lab won the 1st Place of Fall 23 CS SHOWCASE. Congrats Saraswathi Katabattula, Janou Milligan, Tyler Wilkerson!
  • Our paper titled “DeepEn2023: Energy Datasets for Edge Artificial Intelligence” led by Xiaolong has been accepted by Thirty-seventh Conference on Neural Information Processing Systems (NeurIPS ’23) Workshop on Tackling Climate Change with Machine Learning. This achievement aligns with our series of contributions (SEC ’23, CCAI@NeurIPS ’23) aimed at assessing and reducing energy consumption and carbon footprint of on-device deep learning. Kudos to my students for their outstanding work and dedication!
  • Xiaolong was selected as one of the three recipients of the 2023 Martin D. Fraser Graduate Student Conference Travel Award, which is a merit-based accolade that supports CS PhD students at GSU with conference travel. Congratulations!
  • Our paper titled “Real-Time Object Substitution for Mobile Diminished Reality with Edge Computing” led by Hongyu has been accepted by SEC’ 23 Poster/Demo Track.
  • Our paper titled “Unveiling Energy Efficiency in Deep Learning: Measurement, Prediction, and Scoring across Edge Devices” led by Xiaolong has been accepted by The Eighth ACM/IEEE Symposium on Edge Computing (SEC’ 23). SEC is a new top conference in edge computing. The acceptance rate of this year is 25.3% (18 of 71 accepted). This paper presents the first measurement study that uncovers the energy consumption characteristics of on-device deep learning across edge devices. This study results in the creation of three extensive energy datasets for edge devices, covering a wide range of kernels, state-of-the-art DNN models, and popular AI applications. Based on our datasets, we design and implement the first kernel-level energy predictors for edge devices, which achieves an average prediction accuracy of 86.2%. The datasets and code will be released soon at here. Stay tuned!
[Summer 2023]
[Spring 2023] 

Research Highlights

How Do Prompt Variations Affect Energy Consumption in On-Device LLMs? 
[EMNLP’26 Main Conference]

The deployment of Large Language Models (LLMs) on mobile devices is rapidly expanding, making battery life a critical bottleneck for edge AI. While existing efficiency research heavily focuses on model-centric optimizations like quantization or pruning, a fundamental question has remained largely underexplored: how does the design of the prompt itself affect the energy consumption of on-device LLMs? To address this, we conducted the first large-scale empirical study to systematically investigate how two key prompt properties: cognitive load and phrasing pattern, shape the energy behavior of on-device inference. We constructed a novel, controlled dataset to isolate reasoning demands from semantic meaning, and performed fine-grained, phase-level power profiling across diverse models and mobile devices. Our findings reveal a distinct dichotomy: cognitive load primarily drives the energy cost per token, whereas surface-level phrasing patterns dictate energy consumption mainly by altering overall token usage. Furthermore, our analysis shows that smart, model-aware prompt design can significantly push the boundaries of a model’s energy-quality Pareto frontier, offering a powerful new, model-agnostic lever for sustainable on-device LLM.

Deformba: Vision State Space Model with Adaptive State Fusion 
[ICML’26]

State Space Models (SSMs) have emerged as highly efficient, linear-time alternatives to Transformers, but extending them to vision tasks remains challenging due to their reliance on manually designed, fixed scanning methods that disrupt natural 2D spatial structures. Furthermore, standard SSMs lack a natural mechanism for query-based interactions between distinct information streams, a feature that is critical for complex tasks like multi-view 3D fusion. To address these fundamental limitations, we developed Deformba, a novel vision state space framework featuring Context-Adaptive State Fusion (CASF). Deformba decouples the SSM write and read operations, using a single scan to build a shared state memory and an offset predictor to dynamically sample context-relevant spatial evidence. This write-once, adaptive-read-many paradigm naturally supports both 2D spatial modeling and multi-modal query-to-source fusion, functioning similarly to cross-attention, while maintaining the efficiency and linear computational complexity of SSMs. Extensive experiments demonstrate that Deformba achieves strong, state-of-the-art performance across diverse visual perception benchmarks, spanning general 2D tasks like image classification and object detection, to complex 3D tasks like BEV perception.

MamBEV: Enabling State Space Models to Learn Birds-Eye-View Representations 
[ICLR’25]

Automatically generating a bird’s-eye-view (BEV) of an object’s surrounding environment is critical for applications like autonomous driving and advanced driver-assistance systems. These systems rely on integrating signals from multiple cameras to construct a top-down view of the environment. Prominent examples include the BEV systems deployed in Tesla cars. However, many existing methods heavily depend on Transformers, which employ computationally expensive attention mechanisms to learn accurate representations.

In this work, we introduce Spatial Cross Mamba, an innovative approach analogous to standard cross-attention in Transformers. Our method leverages the efficiency of state space models (SSMs) to significantly reduce the computational overhead associated with Transformers, enabling more efficient and scalable BEV systems without compromising representation accuracy.

lm-Meter: Unveiling Runtime Inference Latency for On-Device Language Models
[SEC’25]  Best Paper Finalist

Running LLMs locally on mobile and edge devices promises improved privacy, reliability, and lower communication costs, but remains challenging due to high resource demands and limited visibility into performance-efficiency trade-offs. We present lm-Meter, the first lightweight, online latency profiler for on-device LLM inference. lm-Meter provides fine-grained, real-time phase- and kernel-level latency profiling without auxiliary devices. Implemented on commercial mobile platforms, lm-Meter achieves high accuracy with minimal overhead (2.58% prefill and 0.99% decode throughput loss under the most constrained power settings). Leveraging lm-Meter, we conduct comprehensive empirical studies revealing phase- and kernel-level bottlenecks in on-device LLM inference, quantifying accuracy-efficiency trade-offs, and identifying systematic optimization opportunities. lm-Meter provides unprecedented visibility into the runtime behavior of LLMs on constrained platforms, laying the foundation for informed optimization and accelerating the democratization of on-device LLM systems.

PlatformX: End-to-End Transferable Platform for Energy-Efficient Neural Architecture Search 
[SEC’25]

PlatformX is an end-to-end, fully automated platform for hardware-aware Neural Architecture Search (HW-NAS) targeting energy-efficient deep neural networks on mobile and edge devices. PlatformX integrates four key components: (i) an energy-driven search space that expands conventional NAS design by incorporating energy-critical configurations, enabling exploration of high-efficiency architectures; (ii) a transferable kernel-level energy predictor across devices and incrementally refined with minimal on-device samples; (iii) a Pareto based multi-objective search algorithm that balances energy and accuracy to identify optimal trade-offs; and (iv) a high-resolution runtime energy profiling system that automates on-device power measurement usingexternal monitors without human intervention.

Unveiling Energy Efficiency in Deep Learning: Measurement, Prediction, and Scoring across Edge Devices 
[SEC’23]

Despite the remarkable advances in edge device capabilities such as functionality, computation power, and storage capacity, the limited energy capacity has been the major bottleneck in promoting advanced edge AI applications. For instance, mobile and edge devices are typically powered solely by embedded batteries, so their energy capacity is significantly constrained by form factor requirements, safety considerations, manufacturing costs, and concerns on the environmental impact of the battery technology used.

In this work, we studied the problem of accurate energy measurement, prediction, and understandable scoring of on-device deep learning across edge hardware. We created kernel-, model-, and application-level datasets for on-device deep learning. We designed and implemented the first kernel-level energy predictors on both mobile CPU and GPU. It can provide consistently accurate energy estimation on unseen DNN models.

Mobile Augmented/Virtual Reality (AR/VR) Systems 
[TMC’23, INFOCOM’20]

The goal of this research pillar is to improve system performance, especially latency and energy efficiency, of mobile AR/VR devices. We primarily focus on adapting the system configurations of mobile AR devices to address the trade-offs between user preferences and performance requirements.

We are also recently interested in prototyping systems for emerging AR applications including: collaborative AR, mobile AR with generative AI, AR for scientific data exploration, etc.

Digital Twins for Future Mobility
[VTM’23, IOTJ’22]

The goal of this research pillar is to conduct pioneering study on digital twins for connected and automated vehicles (CAVs). Our vision is to create fair, affordable, and efficient mobility solutions by leveraging digital twins and edge computing.

We are primarily interested in: Visualization of mobility digital twin with NeRF, cooperative perception, security in CAVs, etc.

Research Partners

Sponsors