Autonomous quadruped robot for 5G spectrum coverage measurements

Autonomous quadruped robot for 5G spectrum coverage measurements Martina Callea1, Roberto Masocco2,⋆, Luca Chiaraviglio1, Alessandro Tenaglia2, Alexandru Cretu2, Simone Mattogno2, Daniele Carnevale2 1Dipartimento di Ingegneria Elettronica, Universit` a degli Studi di Roma Tor Vergata 2Dipartimento di Ingegneria Civile e Ingegneria Informatica, Universit` a degli Studi di Roma Tor Vergata Via del Politecnico 1, 00133 Roma, Italy ⋆[email protected] Abstract—This work introduces an autonomous robotic measurement platform, based on a quadruped robot, designed to perform systematic 5G spectrum measurements and area coverage analysis. Fifth-generation (5G) mobile networks have increased the complexity of radio propagation: unlike previous generations, 5G signals show strong spatial and temporal variability, making them difficult to characterize comprehensively with static, manual surveys, thus requiring new measurement approaches. Index Terms—Quadruped robot, Autonomous navigation, 5G I. INTRODUCTION Fifth-generation (5G) networks, employing novel features such as dynamic beamforming and Multi-User MIMO, and characterized by rapidly changing traffic, exhibit high spatial and temporal variability in radio propagation patterns. Traditional measurement campaigns, typically relying on static setups or manual surveys, are costly and time-consuming and do not capture this dynamic behavior with sufficient resolution. These limitations call for new approaches that combine mobility, automation, and adaptability. In this context, this work presents an autonomous quadruped-based platform designed to perform systematic and repeatable spectrum measurements, with the capability to detect both power hotspots and poorly covered areas, adapting the sampling density in response to local radio conditions. The robotic platform has been designed to enable operations even on rugged terrain, such as areas affected by natural disasters where network connectivity is essential to coordinate the rescue effort. II. SYSTEM OVERVIEW The proposed platform, depicted in Figure 1, is based on a Unitree Go2 quadruped robot, adapted to carry professionalgrade instrumentation while autonomously navigating outdoor environments which are often characterized by irregular terrain, physical obstacles, and heterogeneous surface types. To address the specific application considered in this work, the robot has been equipped with a compact payload integrating an Anritsu MS2760A ultra-portable spectrum analyzer, a Quectel RM502Q-AE 5G modem, and an u-blox F9P GNSS-RTK (RealTime Kinematic) receiver: this setup enables synchronized and georeferenced measurements that combine physical-layer This work was supported by European Union through Italian National Recovery and Resilience Plan (NRRP) of NextGenerationEU, partnership on “Telecommunications of the Future” (program “RESearch and innovation on future Telecommunications systems and networks (RESTART)”) under Grant PE00000001. information from channel power acquisitions with network status information provided by the modem. The set of sensors is completed by a Stereolabs ZED X stereo camera providing a real-time depth map of the surrounding area, which is useful for identifying and avoiding obstacles during autonomous navigation [1]. The central hub of the onboard hardware is an Nvidia Jetson Orin AGX computer, whose Graphic Processing Unit (GPU) allows for real-time depth data processing. From the software perspective, the system follows a hierarchical, layered architecture designed for modularity and built using the Robot Operating System 2 (ROS 2) middleware [2]– [4]. At the lowest level, operating system and device drivers ensure direct communication with sensors and instruments. In addition, an integration layer exposes each device (e.g., spectrum analyzer, 5G modem, GNSS-RTK receiver) on structured ROS 2 topics. Next, software in the localization and navigation layers fuses visual odometry, GNSS-RTK, and inertial measurements in Extended Kalman Filters (EKFs) to provide drift-corrected pose estimation, detect obstacles, compute and follow paths. The Go2 robot is interfaced with dedicated Data Distribution Service (DDS) topics [5] bridged to ROS 2, receiving commands and providing status feedbacks and readings of the inertial and odometric sensors. Above this, the mission layer executes the surveying task: finite-state machines orchestrate movement in the area and adaptive maneuvers, such as loitering around power hotspots to gather detailed samples. Finally, a web interface completes the stack, offering real-time visualization of the acquired data. III. AUTONOMOUS CHANNEL POWER MEASUREMENTS Among the narrowband measurement methodologies defined in international EMF assessment guidelines [6], this work adopts channel power as the core metric. In this mode, the spectrum analyzer computes the total received power within a configurable bandwidth together with the corresponding power spectral density (PSD). The channel power is expressed as Pch =Zfc+B/2 fc −B/2 S(f)df, (1) where fc is the channel center frequency, B the integration bandwidth, and S(f) the power spectral density in dBm/Hz . This metric is identified in EMF standards as the primary 2025 I-RIM Conference October 17-19, Rome, Italy ISBN: 9788894580570 10.5281/zenodo.17629786 155 Fig. 1. Go2 robot with 5G measurement hardware and navigation sensors. indicator of exposure levels. Measurements are acquired from the spectrum analyzer via SCPI commands and published on ROS 2 topics. At a higher level, the system is organized around two coordinated finite-state machines (FSMs), one dedicated to navigation, and the other to measurement. The navigation FSM guides the robot through a set of GPS waypoints, ensuring consistent coverage of the measurement area. Although its primary role is trajectory execution, delegated to a navigation module that detects and avoids obstacles on the path, this FSM also adapts the robot’s motion in response to external events such as the detection of a hotspot. In parallel, the measurement FSM continuously monitors the channel power stream. Incoming samples are filtered through an exponentially weighted moving average (EWMA) filter and evaluated using a voting scheme to robustly detect the presence of power hotspots. When one is detected, the measurement FSM publishes a trigger message to the navigation one, signaling that dense sampling is required. Upon reception of this trigger, the navigation FSM temporarily commands the robot to perform a loitering maneuver around the current location, therefore increasing the spatial density of the acquired samples in the hotspot region. Once the maneuver is completed, the navigation FSM resumes its waypoint-based trajectory, and the measurement FSM evaluates distance and decay conditions to resume normal monitoring avoiding hysteresis behaviors. IV. EXPERIMENTAL VALIDATION The proposed platform has been validated through an outdoor campaign on the university campus, leveraging the private 5G standalone (SA) network. In line with EMF assessment guidelines [7], network traffic has been forced using iperf3 to emulate realistic downlink load conditions. The quadruped robot executed predefined waypoint trajectories while channel power samples were continuously acquired, georeferenced, and rendered in real time as heatmaps. This immediate feedback Fig. 2. Channel power heatmap: red areas denote hotspots, while blue spots indicate poorly covered locations. proved essential to identify coverage variations, as shown by the experimental run depicted in Figure 2: in the parking area, three distinct hotspots were detected, clearly visible as regions with higher measurement density due to the adaptive sampling strategy. V. CONCLUSIONS AND FUTURE WORK This work presented an autonomous robotic platform for 5G channel power measurements. The experimental campaign demonstrated reliable and repeatable operation, with the ability to automatically highlight power hotspots through live heatmaps. Future work will extend the platform to zero-span measurements, i.e., time-domain acquisitions at the frequency of the synchronization signal blocks (SSBs). By extrapolating their maximum received power [8] it is possible to derive complementary heatmaps that enrich channel power results with temporal dynamics of beamformed transmissions. REFERENCES [1] L. Bianchi, D. Carnevale, R. Masocco, S. Mattogno, F. Oliva, F. Romanelli, and A. Tenaglia, “Efficient visual sensor fusion for autonomous agents,” in Proc. of the 7th International Conference on Control, Automation and Diagnosis (ICCAD), pp. 01–06, IEEE, May 2023. [2] S. Macenski, T. Foote, B. Gerkey, C. Lalancette, and W. Woodall, “Robot Operating System 2: Design, architecture, and uses in the wild,” Science Robotics, vol. 7, p. eabm6074, May 2022. [3] L. Bianchi, D. Carnevale, F. Del Frate, R. Masocco, S. Mattogno, F. Romanelli, and A. Tenaglia, “A novel distributed architecture for unmanned aircraft systems based on Robot Operating System 2,” IET Cyber-Systems and Robotics, vol. 5, no. 1, p. e12083, 2023. [4] R. Masocco, A unified software architecture for distributed autonomous systems. PhD thesis, Tor Vergata University of Rome, Rome, Italy, 2025. [5] R. Masocco, A. Tenaglia, C. Galperti, D. Carnevale, and S. Galeani, “A distributed real-time diagnostic and control network for the TCV tokamak based on the Data Distribution Service,” Fusion Engineering and Design, vol. 221, p. 115344, Dec. 2025. [6] International Electrotechnical Commission, IEC 62232:2025 - Determination of RF field strength, power density and SAR in the vicinity of base stations for the purpose of evaluating human exposure. Geneva, Switzerland: IEC, 4 ed., 2025. [7] Sistema Nazionale per la Protezione dell’Ambiente, Linee guida per la misura dei campi elettromagnetici generati da sistemi di telecomunicazione in tecnologia 5G, vol. 54. Italy: SNPA, 2025. [8] D. Franci, S. Coltellacci, E. Grillo, S. Pavoncello, T. Aureli, R. Cintoli, and M. D. Migliore, “Experimental Procedure for Fifth Generation (5G) Electromagnetic Field (EMF) Measurement and Maximum Power Extrapolation for Human Exposure Assessment,” Environments, vol. 7, p. 22, Mar. 2020. 156