Lingkun Li 李令昆

Associate Professor and Doctoral Supervisor
School of Software Engineering, Beijing Jiaotong University

I study mobile systems for sensing, localization, and on-device AI. My work uses ambient light, wireless signals, and off-the-shelf hardware to build and evaluate systems in real-world environments.

I received my Ph.D. from Michigan State University, my M.Eng. from Tsinghua University, and my B.Eng. from Beijing Jiaotong University. More about me.

Lingkun Li

Recent News

News archive

I received the Excellent Undergraduate Thesis Advisor, Excellent Class Advisor, and Excellent Lecturer awards from Beijing Jiaotong University.

Our paper on mobile LLM inference was accepted by ACM SIGMETRICS 2027.

I plan to recruit one Ph.D. student in mobile intelligent computing for 2027 entry. Details.

Research

Visible-Light Sensing and Communication

RainbowLight locates a phone using angle-dependent colors from passive tags. It achieved a mean 3D error of 9.6 cm in lights-on tests. LiTag uses the same effect to estimate pose. LuminaLink uses it for position-dependent communication, with a mean bit error rate of 3.0% at the intended position and 52.4% elsewhere in a 28-position test.

Two views of the same four-square optical chip show different colors
The same LuminaLink chip viewed from two positions · INFOCOM 2025, Fig. 7(c–d) · Related NSFC project

Localization and Spatial Intelligence

TransparentLoc uses cellular fingerprints from outdoor trips to recommend ride-hailing pickup points where GPS and Wi-Fi are unavailable. Our two-year DiDi study, described in the TMC extension, covered 50 million orders. In a follow-up study, we used outdoor Wi-Fi fingerprints for indoor localization and reduced median pickup error by 4.35 m compared with Lab2Wild in a shopping-mall experiment.

Shopping mall floor plan marking reference points in blue and measured localization errors in red
Shopping-mall evaluation · IMWUT 2025, Fig. 9(b) · Related NSFC project

Intelligent Transportation

DSETA updates travel-time predictions at two timescales so the model can respond to incidents and holidays while retaining long-term traffic patterns. In an online A/B test on DiDi, mean absolute error fell by 6.62% in Beijing, 0.73% in Wuhan, and 2.40% in Xi’an.

Daily travel-time prediction error for DSETA and the baseline in the Beijing online experiment
Beijing online A/B test · KDD 2026, Fig. 7

Edge AI and Networked Systems

We built PowerBench to measure where energy goes during mobile LLM inference. Shorter CPU–NPU polling intervals cut total energy per token by 30.9–37.8% in our experiments. Most of the savings were on the NPU, consistent with it spending less time active after completing a request.

Shortening the RPC polling interval reduced NPU energy per token by 49.8 to 70.8 percent and total energy per token by 30.9 to 37.8 percent. Throughput changed by −7.2 to +0.1 percent.
Measured ranges across two models and both inference phases, with RPC polling at 5–20 µs versus the 9,999 µs default. Section 8.1 and Fig. 11 · Related Beijing NSF–Xiaomi project

Prospective Ph.D. Students

I plan to recruit one Ph.D. student in mobile intelligent computing for 2027 entry through BJTU’s application and assessment process. Please send your CV to [email protected]. Applicants should be self-motivated and have strong English and programming skills. Programming skills will be assessed during selection.

For an example of this work, see our study of LLM inference on phones, where we measure energy use on each processor and test how CPU–NPU coordination affects it.

计划招收 2027 年入学的移动智能计算方向博士生 1 名,采用申请考核制。请将个人简历发送至 [email protected],遴选安排将统一通过邮件通知。北交大教师主页招生信息

Teaching and Mentoring

In Discrete Mathematics, students write proofs in Coq and submit them through an online lab for immediate feedback. See the teaching paper and the full course list.

I advise undergraduate theses and student projects. A team I advised won first prize in the North China regional round of the 2024 National College Student Software Innovation Competition. BJTU recognized me as an Excellent Undergraduate Thesis Advisor in 2025 and 2026. I also received the Excellent Class Advisor and Excellent Lecturer awards in 2026.

Selected Publications

Full publication list

Professional Service

My reviewing expertise includes mobile sensing and localization, field evaluation of mobile systems, and on-device inference measurement.

My program committee service includes ACM APSys 2025, IEEE GLOBECOM 2025 and 2026, and IEEE ICPADS 2024. I have reviewed for IEEE TMC, ACM TOSN, and ACM IMWUT. Full service record.