Table of Contents

Keynote: HW/SW System co-design

Prof. Rodolfo Pellizzoni, University of Waterloo

  • Embedded is getting more integrated
    • Automotive, multiple processing units in a car
  • Systems have mixed criticality systems
  • Arguing, needing isolation between high/low critical systems
  • Typical way is an embedded hypervisor
    • QNX hypervisor (industry)
    • 2 others??
    • Bao (academia)
  • Problem, lack of timing isolation, leads to lower systems
  • Similar problem in the cloud with co-located VMs
  • Two types of contention
    • (1) spacial, limited resources
    • (2) temporal, limited time
  • Adjust control policy based on the current state of the system
    • Can be done in both hw/sw
    • Arguing that co-design increases performance
  • Original work (MemGuard) had enforcement implemented in sw
    • Slowdwon from procesing on the core
    • Coarse grained control (on/off)
  • MemPool had an external HPC on a side core with a speedup
    • Find grain control to week at more fine grained control
  • MemCoRe implemented in a FPGA with hardware obsv. ctrl. and encorement in sw
  • Proposing having another lane depending on the criticality of core
  • Specification is on the arch but depends on the implementation
  • Proposing regulating the arbitrator rather than lane
  • Dynamical disable speculative requests depending on load
    • Implemented in DAMA
  • ASIL guarantees and ISO 26262
  • Accelerators also need to be controlled, highest memory bandwidth
  • Takeaways
    • Integrated mixed-critical SoC need performance isolation
    • Shared HW breaks isolation
    • HW support improves resource management
    • Heterogeneous HW for hetero requirements
    • Lack good SW for controlling HW

SenSys: Sky-High Buddies: Drones & Space Friends

WildFiT: Autonomous In-Situ Model Adaptation for Resource-Constrained IoT Systems

Mohammad Mehdi Rastikerdar, Jin Huang, Hui Guan, Deepak Ganesan (University of Massachusetts Amherst)

WISP: Printable Graphene-Based Wearables for Force-Based Micro-Gesture Recognition

Zhenyu Lei, Xiaomeng Liu, Quan Zhang, VP Nguyen, Jun Yao, Deepak Ganesan (University of Massachusetts Amherst)

  • How to detect mm-scale interactions
  • Measure skin and tendon deformation
  • PI: seems like kapton tape?
  • LIG: Laser-induced graphene
    • Allow for the printing of sensors
  • Need to tune the laser to correctly etch graphene
  • FEM to reduce cross-talk and prevent tearing
  • Attached to the underside of the wrist
    • Or other types of objects??

VYRE: Low-Burden and Robust Oscillometric Ring-Based System for Frequent Blood Pressure Monitoring

Amirmohammad Radmehr, Shamanth Kuthpadi Seethakantha, Abdul Aziz, Quang Trung Tran, Aryan Nair, William Saulnier, Deepak Ganesan, Phuc Nguyen (University of Massachusetts Amherst)

  • Look at monitoring blood pressure for hypertension
  • Mechanical oscillation to infer blood pressure
  • Have v2 prototype with 2133 patients in hospital
  • 300-400 mW power consumption

Med2ECG: Medical-Guided BCG-To-ECG Reconstruction for Diverse Populations

Lin Chen (HKUST (GZ)); Yandao Huang, Chenggao Li (HKUST); Jun Chen, Shuxin Zhong, Minghui Qiu, Chunzhen Guo, Yi Wang (HKUST (GZ)); Qian Zhang (HKUST); Kaishun Wu (HKUST (GZ))

  • Using PDMS flex fibers
    • Same material that was used for biomemetic leaf wetness sensor
  • Didn’t fully understand the work

EarPCG: Recovering Heart Sounds from in-Ear Audio via Physics-Informed Neural Network

Junyi Zhou, Yiyi Zhang (Huazhong University of Science and Technology); Henglin Pu (Purdue University); Peng Guo (Huazhong University of Science and Technology); Tianyue Zheng (Southern University of Science and Technology); Chao Cai (Huazhong University of Science and Technology); Jun Luo (Nanyang Technological University)

  • In-ear audio carry heartbeat information
  • Reconstruct PCG from in-ear audio
  • NN methods for relating the two measurements
  • PINN???
    • Method of incorperating equations in NN model?
  • Build physical testbed

NutriEar: Robust Nutrition-Aware Food Classification from In-Ear Acoustic Signals

*Zoey Xiaochen Tan (University of Cambridge); Yang Liu (Florida State University); Kayla-Jade Butkow, Cecilia Mascolo (University of Cambridge)

  • Create taxonomy based on texture
  • Earable device that takes sounds and converts it to the texture
  • Signal changes over time as you chew
  • ML methods to group different groups
  • Unqiue approach to detecting bytes

MRI-Grade Photoplethysmography Using Bundled Fiber Optics for Contactless Heart Rate Monitoring and Real-Time Gating*

Tommaso Polonelli, Sébastien Emery, Bianca Müller, Ivan Simeonov, Marco Giordano, Michele Magno, Sebastian Kozerke (ETH Zürich)

  • Motion artifacts, breathing, heart rate affect MRI images
  • Cannot use metals in MRI
  • ECG used to trigger the MRI immagery
  • Use fiber optic to detect PPG
  • 2.8ms processing time, lower than heartbeat period
  • Contactless PPG

Short Paper: EarSleeve: Transforming Everyday Earphones into a 12-Lead ECG Sensing Platform

Junxi Xia, Doğaç Eldenk, Hongjun Xu (Northwestern University); Yang Liu (Florida State University); Stephen Xia (Northwestern University)

  • Wasn’t interested

WristSense: Sensing Hidden Wrist Strain in Routine Activities via Inertial Tokenization and LLM-Based Feedback

Garvit Chugh, Ananya Mondal (Indian Institute of Technology Jodhpur); Sandip Chakraborty (Indian Institute of Technology Kharagpur); Suchetana Chakraborty (Indian Institute of Technology Jodhpur)

  • Able to get wrist and finger angles!

Short Paper: Towards Real-Time ECG and EMG Modeling on μNPUs

Josh Millar (Imperial College London); Ashok Samraj Thangarajan (Nokia Bell Labs); Soumyajit Chatterjee (Brave Software, University of Cambridge); Hamed Haddadi (Imperial College London)

  • uNPUs have les than 100 mW of power draw
  • Want to put foundational models on microcontroller hardware.

From a Point to Hundreds: Embracing LiDAR on Commodity Smartphones for Fine-Grained Pulmonary Function Sensing

Xuefu Dong (The University of Tokyo); Wenwei Li (Peking University); Minhao Cui (Seoul National University); Zilong Wang (Microsoft Research Asia); Lupeng Zhang (Nanyang Technological University); Akihito Taya, Yuuki Nishiyama, Kaoru Sezaki (The University of Tokyo); Lili Qiu (Microsoft Research Asia, UT Austin); Jie Xiong (Nanyang Technological University)

  • Wasn’t interested

NeuroPath: Practically Adopting Motor Imagery Decoding through EEG Signals

Jiani Cao, Kun Wang (City University of Hong Kong); Yang Liu (Florida State University); Zhenjiang Li (City University of Hong Kong)

  • Wasn’t interested

Short Paper: WearBCI Dataset: Understanding and Benchmarking Real-World Wearable Brain-Computer Interfaces Signals

Haoxian Liu, Hengle Jiang, Lanxuan Hong, Xiaomin Ouyang (Hong Kong University of Science and Technology)

  • Wasn’t interested

SenSys: Sky-High Buddies: Drones & Space Friends

Boyi Liu (HKUST & University College London); Qianyi Zhang (University College London); Qiang Yang (University of Cambridge), Jianhao Jiao, Chauhan Jagmohan, Dimitrios Kanoulas (University College London)

  • Existing work is closed source and stationary, not open sourced
  • https://starlinkrobot.github.io
  • Robot to collect position and network information, LIDAR, sky cover
  • Velocity has an impact on link quality
  • Sky obstructions causes RTT instability

Exploring LEO Satellite Networks for Continuous Monitoring and Dynamic Tracking

Lang Wei,Ruichen Li, Baodong Chen, Qifan Yang, Yufan Wu, Ting Zhu (The Ohio State University)

  • Looking to enable continuous monitoring of land from satellites
  • Can reduce the gap between measurements
  • Coordinates imagery across space and time
  • Exploratory work for using dense LEO satellites for ground sensing

SERENADE: A Digital Twin Emulator for LEO Satellite Networking At-Scale

Roberto Chamorro Martinez, N. Cameron Matson, Karthik Sundaresan (Georgia Institute of Technology)

  • LARGE scale (greather than 500000 nodes)
  • Uses threads instead of containers
  • Can use external client/server containers
  • Scales linearly in computation complexity in terms of satellites
  • Lower start and stop times

Enabling Near-realtime Remote Sensing via Satellite–Ground Collaboration of Large Vision–Language Models

Zihan Li, Jiahao Yang, Yuxin Zhang, Zhe Chen, Yue Gao (Fudan University)

  • German satellite data archive added over 6000 PB
    • good resource for future work
  • Limitations in the size for large scale models
  • Two models, one on satellite and one on ground
  • Interesting problem with online inference of data, and trusting the result

LaSen: Low-Altitude Drone Sensing with 5G-NR Signals

Qian Yang (Southern University of Science and Technology, Peng Cheng Laboratory); Yongtao Dai, Mingrui Li, Qianyi Huang, Xu Chen (Sun Yat-sen University); Jin Zhang (Southern University of Science and Technology), Guochao Song (China Academy of Information and Communications Technology); Qian Zhang (The Hong Kong University of Science and Technology); Xiaofeng Tao (Beijing University of Posts and Telecommunications, Peng Cheng Laboratory)

  • ISAC for UAV sensing
  • 5G reference is too sparse to track high speed UAV
  • Looking to do it without extra spectrum cost

Count Every Rotation and Every Rotation Counts: Exploring Drone Dynamics via Propeller Sensing

Xuecheng Chen (Tsinghua University); Jingao Xu (The University of Hong Kong); Wenhua Ding, Haoyang Wang, Xinyu Luo (Tsinghua University), Ruiyang Duan, Jialong Chen (Meituan Academy of Robotics); Xueqian Wang, Yunhao Liu, Xinlei Chen (Tsinghua University)

  • https://eventpro25.github.io/EventPro
  • Determine, what the drone intends to and what the drone is currently doing
  • Use the propeller rotation speed to infer internal flight commands
  • Neurotrophic camera to generate events
  • Autofocus and tracking for the drone

(Interesting) GreenScatter: Through-Canopy Soil Moisture Sensing with UAV-Mounted Radar

Luke Jacobs, Ishfaq Aziz, Benhao Lu (University of Illinois Urbana-Champaign); Alireza Tabatabaeenejad (The Aerospace Corporation); Mohamad Alipour, Elahe Soltanaghai (University of Illinois Urbana-Champaign)

  • Separate ground reflect from the canopy reflection
  • Attenuation of the canopy changes over time
  • LiDAR based canopy characterization
  • Compare measure vs predicted RCS (radiative transfer model)
  • Pulsed radar for time domain RCS conversion

ARC: Accurate, Real-Time, and Scalable Multi-Vehicle Cooperative Perception

Kaleem Nawaz Khan, Fawad Ahmad (Rochester Institute of Technology)

  • https://github.com/nsslofficial/ARC
  • Combined point clouds between vehicles
  • Aligning point clouds
  • Find overlapping regions and align based on that subset
  • Intelligent delta way of sending missing points

CARTS: Cooperative and Adaptive Resource Triggering and Stitching for 5G ISAC

Cheng Jiang, Yihe Yan (University of New South Wales); Yanxiang Wang (Shandong University); Jiawei Hu, Chun Tung Chou, Wen Hu (University of New South Wales)

  • Fusing DMRS and SRS
  • Standard compliance
  • Way too technical for me

EventEye: Towards High-Frequency Perception Enhancement for Autonomous Vehicles Using Infrastructure-Mounted Event Cameras

Jingfei Xia (The Chinese University of Hong Kong); Yuze He (Carnegie Mellon University); Chen Bian, Zhenyu Yan, Guoliang Xing (The Chinese University of Hong Kong)

  • Drive to make light posts “smart”
  • Event driven camera to enhance autonomous vehicle perception
    • Very similar to above work
  • Converting between event camera and vehicle LiDAR
  • BEV map??
    • Looks like prior work

COACH: Adaptive Robust Human-Robot Collaboration for Efficient Smart Manufacturing

Hui Wang (Harbin Engineering University); Liekang Zeng (The Chinese University of Hong Kong); Zhiwen Yu (Harbin Engineering University, Northwestern Polytechnical University); Yao Zhang (Northwestern Polytechnical University); Di Duan, Mu Yuan (The Chinese University of Hong Kong); Bin Guo (Northwestern Polytechnical University); Guoliang Xing (The Chinese University of Hong Kong)

  • Scheduling human/robotic tasks
  • Done in simulation on a small IoT assembly line
  • Didn’t understand how policies would be implemented

Defending Autonomous Driving Perception against Adversarial Object-Based Attacks via Motion Planning

Zihao Liu, Yan Zhang (Iowa State University); Yi Zhu (Wayne State University); Lu Su (Purdue University); Chunming Qiao (SUNY at Buffalo); Chenglin Miao (Iowa State University)

  • NN object detection is opne to attacks
  • Vehicle-hiding attack where adversarial objects are placed around the car

Been There, Scanned That: Nostalgia-Driven Point Cloud Compression for Self-Driving Cars

Ali Khalid, Jaiaid Mobin (Rochester Institute of Technology); Sumanth Rao Appala (University of Pennsylvania); Avinash Maurya (Rochester Institute of Technology); Stephany Berrio Perez (The University of Sydney); M. Mustafa Rafique, Fawad Ahmad (Rochester Institute of Technology)

  • https://github.com/nsslofficial/DejaView
  • Has large data requirements 4 PB per year
  • Compress point cloud data
    • Shared points between frames
    • Shared points for fixed objects (buildlings)
  • Cascaded difference between point clouds

Deadlines for next year

First deadline is 3 weeks earlier

  • June 5th deadline??