Michele Magno

Last Name

Magno

First Name

Michele

ORCID

0000-0003-0368-8923

Organisational unit

01225 - D-ITET Zentr. f. projektbasiertes Lernen / D-ITET Center for Project-Based Learning

Show all metadata

Search Results

Publications 1 - 10 of 148

A Cost-effective Occupancy Estimation System for Energy-efficient Buildings in Africa
Adordie, Godwin; Kyalo Muuo, Denis; Schilk, Philipp; et al. (2025)
2025 IEEE Sensors Applications Symposium (SAS)
Intelligent energy management in buildings is essential to optimize the use of Heating, Ventilation, and Air Conditioning (HVAC) systems, reducing their power consumption and operational costs. Traditional approaches to energy management in buildings rely on fixed schedules, static thresholds, or simple sensor-based control, which are often not aware of momentary power needs and thus fail to optimize energy use effectively. On the other hand, efficient management is crucial, especially in Africa, where high energy costs and unreliable or limited electricity supply necessitate cost-effective energy-saving solutions.This paper presents a robust, cost-effective, easy-to-install intelligent occupancy detection system based on widely adopted and cost-efficient HC-SR04 ultrasonic sensors. One of the main contributions of this paper is a quantized and lightweight machine learning algorithm designed to match the stringent requirements of the ARM Cortex-M4 cores to process the data locally, eliminating the need for a centralized control center. A comprehensive dataset of over 300 events from all combinations of entering and exiting groups of up to three people has been collected, annotated, and utilized for algorithm training and evaluation. Experimental results demonstrate a detection accuracy of 92% in occupancy change for up to three people passing through a doorway simultaneously. Quantizing the model reduced its size by 75% (from 168KB to 42KB) and RAM usage by 50% (from 32KB to 15KB), with no measurable drop in performance, enabling efficient deployment across resource-constrained microcontrollers, including direct in-sensor processing architectures. In contrast to computationally intensive, privacy-sensitive vision-based systems, our approach exploits ultrasonic sensing combined with machine learning to embed AI directly onboard resource-limited sensors, offering a lightweight, privacy-preserving, and cost-efficient occupancy detection solution.
BodySense: An Expandable and Wearable-Sized Wireless Evaluation Platform for Human Body Communication
Schulthess, Lukas; Mayer, Philipp; Vogt, Christian; et al. (2025)
2025 IEEE International Instrumentation and Measurement Technology Conference (I2MTC)
Wearable, wirelessly connected sensors have become a common part of daily life and have the potential to play a pivotal role in shaping the future of personalized healthcare. A key challenge in this evolution is designing long-lasting and unobtrusive devices. These design requirements inherently demand smaller batteries, inevitably increasing the need for energy-sensitive wireless communication interfaces. Capacitive Human Body Communication (HBC) is a promising, power-efficient alternative to traditional RF-based communication, enabling point-to-multipoint data and energy exchange. However, as this concept relies on capacitive coupling to the surrounding area, it is naturally influenced by uncontrollable environmental factors, making testing with classical setups particularly challenging. This work presents a customizable, wearable-sized, wireless evaluation platform for capacitive HBC, designed to enable realistic evaluation of wearable-to-wearable applications. Comparative measurements of channel gains were conducted using classical grid-connected- and wireless Data Acquisition (DAQ) across various transmission distances within the frequency range of 4 MHz to 64 MHz and revealed an average overestimation of 18.15 dB over all investigated distances in the classical setup.
WakeMod: A 6.9 µW Wake-Up Radio Module with –72.6 dBm Sensitivity for On-Demand IoT
Schulthess, Lukas; Cortesi, Silvano; Magno, Michele (2025)
2025 10th International Workshop on Advances in Sensors and Interfaces (IWASI)
Large-scale Internet of Things (IoT) applications, such as asset tracking and remote sensing, demand multi-year battery lifetimes to minimize maintenance and operational costs. Traditional wireless protocols often employ duty cycling, introducing a tradeoff between latency and idle consumption – both unsuitable for event-driven and ultra-low power systems.A promising approach to address these issues is the integration of always-on wake-up radios (WuRs). They provide asynchronous, ultra-low power communication to overcome these constraints.This paper presents WakeMod, an open-source wake-up transceiver module for the 868 MHz ISM band. Designed for easy integration and ultra-low power consumption, it leverages the −75 dBm sensitive FH101RF WuR. WakeMod achieves a low idle power consumption of 6.9 µW while maintaining responsiveness with a sensitivity of −72.6 dBm. Reception of a wake-up call is possible from up to 130 m of distance with a −2.1 dBi antenna, consuming 17.7 µJ with a latency below 54.3 ms. WakeMod’s capabilities have further been demonstrated in an e-ink price tag application, achieving 7.17 µW idle consumption and enabling an estimated 8-year battery life with daily updates on a standard CR2032 coin cell. WakeMod offers a practical solution for energy-constrained, long-term IoT deployments, requiring low-latency, and on-demand communication.
Embedded 2D LiDAR-Based Person Tracking for Safe Navigation in Assistive Autonomous Robots
Plozza, Davide; Marty, Steven; Scherrer, Cyril; et al. (2025)
2025 10th IEEE International Workshop on Advances in Sensors and Interfaces (IWASI)
This paper presents a fully embedded real-time person tracking pipeline for assistive quadrupedal robots supporting safe navigation for visually impaired users. Our approach combines a deep learning-based 2D LiDAR person detector with a lightweight multi-object tracker and integrates it into a Guide Dog Robot (GDR) navigation framework. A novel detection post-processing scheme is proposed, reducing detector latency by 52.38% compared to state-of-the-art voting-based methods while preserving accuracy. The improved latency enables the entire pipeline to operate reliably at 20 Hz on a resource-constrained mobile robotic embedded platform based on the NVIDIA Jetson Xavier NX. The experimental setup shows that our system tracks dynamic obstacles and continuously localizes the user holding the robot’s handle, enabling a dynamic safety footprint for proactive collision avoidance. Under the tested setup, the optimal configuration achieves a MOTA of 83.27% and a user tracking RMSE below 0.2 m on two custom datasets recorded with motion-capture ground truth. Real-world navigation experiments in indoor environments demonstrate effective collision prevention and smooth corrective maneuvers when the user drifts from the default following position. The modular design of the detection, tracking, and planning components ensures flexibility and ease of integration into other robotic platforms. This work contributes a scalable and efficient tracking and navigation solution for human-aware mobile robots operating in dynamic environments, supporting safer human-robot interaction in assistive contexts.
R-CARLA: High-Fidelity Sensor Simulations with Interchangeable Dynamics for Autonomous Racing
Brunner, Maurice; Ghignone, Edoardo; Baumann, Nicolas; et al. (2025)
2025 IEEE Intelligent Vehicles Symposium (IV)
Autonomous racing has emerged as a crucial testbed for autonomous driving algorithms, necessitating a simulation environment for both vehicle dynamics and sensor behavior. Striking the right balance between vehicle dynamics and sensor accuracy is crucial for pushing vehicles to their performance limits. However, autonomous racing developers often face a trade-off between accurate vehicle dynamics and high-fidelity sensor simulations. This paper introduces R-CARLA, an enhancement of the CARLA simulator that supports holistic full-stack testing, from perception to control, using a single system. By seamlessly integrating accurate vehicle dynamics with sensor simulations, opponents simulation as NPCs, and a pipeline for creating digital twins from real-world robotic data, R-CARLA empowers researchers to push the boundaries of autonomous racing development. Furthermore, it is developed using CARLA's rich suite of sensor simulations. Our results indicate that incorporating the proposed digital-twin framework into R-CARLA enables more realistic full-stack testing, demonstrating a significant reduction in the Sim-to-Real gap of car dynamics simulation by 42% and by 82% in the case of sensor simulation across various testing scenarios.
WakeLoc: An Ultra-Low Power, Accurate and Scalable On-Demand RTLS using Wake-Up Radios
Cortesi, Silvano; Vogt, Christian; Magno, Michele (2025)
IEEE INFOCOM 2025 - IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS)
For future large scale robotic moon missions, the availability of infrastructure-less, cheap and low power real-time locating systems (RTLSs) is critical. Traditional RTLS face significant trade-offs between power consumption and localization latency, often requiring anchors to be connected to the power grid or sacrificing speed for energy efficiency. This paper proposes WakeLoc, an on-demand RTLS based on ultra-wideband (UWB), enabling both low-latency and ultra-low power consumption by leveraging UWB wake-up radios (WuRs). In WakeLoc, tags independently start a localization procedure by sending a wake-up call (WuC) to anchors, before performing the actual localization. Distributed tags equipped with WuRs listen to the WuC and use passive listening of the UWB messages to determine their own position. Experimental measurements demonstrate that the localization accuracy in a 2D setup achieves less than 12.9cm error, both for the active and the passive tag. Additional power simulations based on real-world measurements were performed in a realistic environment, showing that anchors can achieve a power consumption as low as 15.53{\mu}W while the RTLS performs one on-demand localization per minute for 5 tags, thus operate up to 5.01 years on a single coin cell battery (690mWh).
FPGA-Accelerated Hybrid Lossless and Lossy Compression for Next-Generation Portable Optoacoustic Platforms
Villani, Federico; Mathys, Sevrin; Özsoy, Çağla; et al. (2024)
2024 IEEE Ultrasonics, Ferroelectrics, and Frequency Control Joint Symposium (UFFC-JS)
Optoacoustic (OA) imaging combines optical excitation and ultrasound beamforming to render images of deep tissues with functional contrast and high spatial and temporal resolution; however, its high data rates and large channel count pose significant challenges to developing portable OA systems with limited bandwidth and computational resources. In this context, on-device data compression is required to maximize the information throughput. Traditional lossless compression (LLC) preserves image fidelity but is characterized by non-constant compression ratios that can result in bandwidth saturation. On the other hand, lossy compression (LC) achieves higher and controlled compression at the expense of image quality. This work presents an FPGA-based hybrid compression algorithm that dynamically combines LLC and LC to adapt to bandwidth constraints in real-time, ensuring consistent data transmission while preserving critical image features. Implemented on a Xilinx Kria K26 FPGA, the algorithm can compress 400 MSamples/s, processing multiple analog-to-digital converter channels per compute unit with low FPGA resource utilization (27.1k registers and 35.1 Look-up tables). The hybrid algorithm achieves an average compression ratio of 5.3 ± 0.3 in LLC mode and up to 11 ± 0.1 when the recovery mode activates LC, validated on 6,000 frames of a human OA dataset. Image quality comparisons on human vasculature datasets show that key anatomical details are preserved even during LC. The proposed approach prevents bandwidth saturation without significant hardware overhead, supporting the development of next-generation portable OA imaging systems with efficient and low-cost I/O interfaces.
Residual Koopman Model Predictive Control for Enhanced Vehicle Dynamics with Small On-Track Data Input
Fu, Yonghao; Hu, Cheng; Xiong, Haokun; et al. (2025)
In vehicle trajectory tracking tasks, the simplest approach is the Pure Pursuit (PP) Control. However, this single-point preview tracking strategy fails to consider vehicle model constraints, compromising driving safety. Model Predictive Control (MPC) as a widely adopted control method, optimizes control actions by incorporating mechanistic models and physical constraints. While its control performance critically depends on the accuracy of vehicle modeling. Traditional vehicle modeling approaches face inherent trade-offs between capturing nonlinear dynamics and maintaining computational efficiency, often resulting in reduced control performance. To address these challenges, this paper proposes Residual Koopman Model Predictive Control (RKMPC) framework. This method uses two linear MPC architecture to calculate control inputs: a Linear Model Predictive Control (LMPC) computes the baseline control input based on the vehicle kinematic model, and a neural network-based RKMPC calculates the compensation input. The final control command is obtained by adding these two components. This design preserves the reliability and interpretability of traditional mechanistic model while achieving performance optimization through residual modeling. This method has been validated on the Carsim-Matlab joint simulation platform and a physical 1:10 scale F1TENTH racing car. Experimental results show that RKMPC requires only 20% of the training data needed by traditional Koopman Model Predictive Control (KMPC) while delivering superior tracking performance. Compared to traditional LMPC, RKMPC reduces lateral error by 11.7%-22.1%, decreases heading error by 8.9%-15.8%, and improves front-wheel steering stability by up to 27.6%. The implementation code is available at: https://github.com/ZJU-DDRX/Residual Koopman.
On-device Learning of EEGNet-based Network For Wearable Motor Imagery Brain-Computer Interface
Bian, Sizhen; Kang, Pixi; Moosmann, Julian; et al. (2024)
ISWC '24: Proceedings of the 2024 ACM International Symposium on Wearable Computers
Electroencephalogram (EEG)-based Brain-Computer Interfaces (BCIs) have garnered significant interest across various domains, including rehabilitation and robotics. Despite advancements in neural network-based EEG decoding, maintaining performance across diverse user populations remains challenging due to feature distribution drift. This paper presents an effective approach to address this challenge by implementing a lightweight and efficient on-device learning engine for wearable motor imagery recognition. The proposed approach, applied to the well-established EEGNet architecture, enables real-time and accurate adaptation to EEG signals from unregistered users. Leveraging the newly released low-power parallel RISC-V-based processor, GAP9 from Greeenwaves, and the Physionet EEG Motor Imagery dataset, we demonstrate a remarkable accuracy gain of up to 7.31% with respect to the baseline with a memory footprint of 15.6 KByte. Furthermore, by optimizing the input stream, we achieve enhanced real-time performance without compromising inference accuracy. Our tailored approach exhibits inference time of 14.9 ms and 0.76 mJ per single inference and 20 us and 0.83 uJ per single update during online training. These findings highlight the feasibility of our method for edge EEG devices as well as other battery-powered wearable AI systems suffering from subject-dependant feature distribution drift.
Robustness Evaluation of Localization Techniques for Autonomous Racing
Lim, Tian Yi; Ghignone, Edoardo; Baumann, Nicolas; et al. (2024)
2024 Design, Automation & Test in Europe Conference & Exhibition (DATE)
This work introduces SynPF, an MCL-based algorithm tailored for high-speed racing environments. Benchmarked against Cartographer, a state-of-the-art pose-graph SLAM algorithm, SynPF leverages synergies from previous particle-filtering methods and synthesizes them for the high-performance racing domain. Our extensive in-field evaluations reveal that while Cartographer excels under nominal conditions, it struggles when subjected to wheel-slip-a common phenomenon in a racing scenario due to varying grip levels and aggressive driving behaviour. Conversely, SynPF demonstrates robustness in these challenging conditions and a low-latency computation time of 1.25 ms on on-board computers without a GPU. Using the F1TENTH platform, a 1:10 scaled autonomous racing vehicle, this work not only highlights the vulnerabilities of existing algorithms in high-speed scenarios, tested up until 7.6ms(-1), but also emphasizes the potential of SynPF as a viable alternative, especially in deteriorating odometry conditions.

Publications 1 - 10 of 148

Michele Magno

Last Name

First Name

ORCID

Organisational unit

Filters

Search Results