Safety-guaranteed and Goal-oriented Semantic Sensing, Communication, and Control for Robotics

Wenchao Wu, Shutong Chen, Wenjie Liu, Zhibo Pang, Yansha Deng · 2026

Read Safety-guaranteed and Goal-oriented Semantic Sensing, Communication, and Control for Robotics by Wenchao Wu, Shutong Chen, Wenjie Liu, Zhibo Pang, Yansha Deng (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-13502.txt.

Introduction

Wenchao Wu and co-authors study safety-aware goal-oriented semantic sensing, communication, and control for robotics. Wireless links to edge servers can flood with data, adding latency that degrades robot tasks. Goal-oriented semantic communication sends only goal-relevant data, but it often overlooks safety needs. The authors propose a co-design that treats safety across sensing, communication, and control as one closed loop. They set out an architecture, use cases, and safety requirements. A UAV tracking case study more than doubled safety and success rates.

Catalog record

CREATORWenchao Wu, Shutong Chen, Wenjie Liu, Zhibo Pang, Yansha Deng
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE36114
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.13502

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Safety-guaranteed and Goal-oriented Semantic Sensing, Communication, and Control for Robotics Wenchao Wu, Shutong Chen, Wenjie Liu, Zhibo Pang, Yansha Deng Abstract Wirelessly-connected robotic system empowers robots with real-time intelligence by leveraging remote computing resources for decision-making. However, the data exchange between robots and base stations often overwhelms communication links, introducing latency that undermines real-time response. To tackle this, goal-oriented semantic communication (GSC) has been introduced into wirelessly-connected robotic systems to extract and transmit only goal-relevant semantic representations, enhancing communication efficiency and task effectiveness. However, existing GSC approaches focused primarily on optimizing effectiveness metrics while overlooking safety requirements, which should be treated as the top priority in real-world robotic systems. To bridge this gap, we propose safety-guaranteed and goal-oriented semantic communication for wirelessly-connected robotic system, aiming to maximize the robotic task effectiveness subject to practical operational safety requirements. We first summarize the general safety requirements and effectiveness metrics across typical robotic tasks, including robot arm grasping, unmanned aerial vehicle (UAV)-assisted tasks, and multi-robot exploration. We then systematically analyze the unique safety and effectiveness challenges faced by wirelessly-connected robotic system in sensing, communication, and control. Based on these, we further present potential safety-guaranteed and goal-oriented sensing, communication, and control solutions. Finally, a UAV target tracking case study validates that our proposed GSC solutions can significantly improve safety rate and tracking success rate by more than 2 times and 4.5 times, respectively. I Introduction Wirelessly-connected robotic system has emerged as a new paradigm for enabling intelligent, responsive, and decentralized robotic systems. However, existing wireless networks cannot meet the stringent timeliness requirements of bidirectional communication between robots and base stations, especially in large-scale deployments. On the one hand, the uplink must support the real-time transmission of bandwidth-intensive multi-modal sensory data, where limited wireless capacity can degrade the freshness and completeness of the sensory information, leading to delayed or suboptimal decision-making. On the other hand, the downlink transmission of control and command (C&C) demands timely transmission with high update frequency and reliability, where any packet loss or congestion might impair the robot’s safe control. To address this, goal-oriented semantic communication (GSC) has been introduced as a promising communication paradigm [ 15 ] , which extracts and transmits the significance or usefulness of the raw data with respect to the data exchange goal. For uplink sensory data, recent research has expanded GSC design for different data types in various tasks [ 9 , 12 ] . In [ 9 ] , the scene graphs were extracted from unmanned aerial vehicle (UAV)-captured images as semantic representations of the original image, to preserve key geometric relationships while enabling accurate downstream visual question answering task. In [ 12 ] , a GSC framework was proposed for point cloud-based avatar reconstruction in the Metaverse, where avatar skeleton nodes are ranked based on its importance for selective transmission. For downlink C&C data, prior work has utilized GSC to selectively transmit most useful downlink packets [ 13 , 3 ] , thereby reducing unnecessary communication without compromising downstream control objectives. The authors in [ 13 ] introduced the value and age of information to quantify the semantic importance and freshness of C&C packets and drop less informative packets. Another study [ 3 ] developed a deep reinforcement learning (DRL) algorithm to perform temporal selection on C&C packets, which selectively discards packets that contribute little to the downstream control task. Nevertheless, real-world robotic applications demand more than just task effectiveness, they must also meet stringent safety requirements. According to industrial robotics standards [ 7 , 2 ] , both the US and the EU have established formal regulations that define specific safety requirements for different types of robots operating in various scenarios. For instance, in UAV target tracking, precise tracking demands the UAV stay close to the target, but maintaining a distance of less than 3 meters significantly increases collision risk. In addition to published standards, some studies in the robotics community have incorporated safety as part of the optimization objective or explicit constraints. For example, collision probability has been widely adopted as an optimization objective alongside shortest path planning in autonomous driving applications [ 5 ] , while for collaborative robots, internal joint force has been used as a key safety criteria [ 11 ] to trigger emergency stop for preventing human injury. Unfortunately, existing works in the communication community on robotic communication remain focused on optimizing traditional metrics, such as throughput [ 12 ] and mean squared error (MSE) [ 13 ] , without realizing that safety is a fundamental effectiveness metric across all robotic applications that should be prioritized above all other metrics. Moreover, another unresolved yet critical issue in both the communication and robotics communities is the lack of a unified framework that jointly considers sensing, communication, and control in a closed-loop manner. The safety and effectiveness of control rely on the quality of communication and the accuracy of sensed data; the control decisions in turn influence subsequent sensing and communication; while the type of sensed data and the requirements of the downstream control task affect the communication frequency and transmission priorities. Although recent studies [ 9 , 12 , 13 , 3 ] have explored the pairwise of two components, they typically neglect the tight coupling among all three components. Consequently, one aspect is inevitably sacrificed or insufficiently addressed, leading to suboptimal efficiency or compromising system safety. Motivated by the above, this work systematically investigates how to enable safety-guaranteed and goal-oriented semantic sensing, communication, and control co-design in wirelessly-connected robotic system, aiming to maximize the task effectiveness while satisfying safety requirements. To the best of our knowledge, this is the first work to extend communication design beyond task effectiveness by explicitly incorporating practical safety requirements into the communication objective. Our main contributions include: • We summarize the safety requirements and effectiveness metrics for wirelessly-connected robotic systems in representative industrial robotic tasks. • We identify the key challenges of wirelessly-connected robotic system during sensing, communication and control. To tackle these challenges, we present novel safety-guaranteed and goal-oriented sensing, communication, and control solutions, aiming to maximizing task effectiveness while satisfying safety requirements. • We validate our proposed GSC solutions in a UAV target tracking use case. Compared with baseline, our solutions improve the safety rate by more than 2 times and tracking success rate by more than 4.5 times, respectively. TABLE I: Commonly-used sensors in wirelessly-connected robotic system Robot Embeded Sensor Sensor Type Data Type Devices Transmission Frequency Data Load Tactile Sensor Haptic Tactile Array Sensor 100-1000 Hz Low-Medium Torque Sensor Force/Torque F/T Sensor (e.g., ATI Mini40) 500-2000 Hz Low-Medium Joint Encoder Joint Angles/Velocities Optical or Magnetic Encoder 500-1000 Hz Low Inertial Measurement Unit (IMU) Acceleration, Angular Velocity MPU6050, Xsens MTi 100-1000 Hz Low-Medium Magnetometer Heading HMC5883L, BNO055 10-100 Hz Low External Sensor Sensor Type Data Type Devices Transmission Frequency Data Load Depth Camera Depth Image Azure Kinect Sensor 10-60 Hz Medium-High LiDAR Sensor 3D Point Cloud Velodyne, Ouster LiDAR 5-20 Hz High Ultrasonic Sensor Distance HC-SR04, MaxBotix 1-50 Hz Low UWB Sensor Radio Signal Decawave DW1000 10-100 Hz Low Wi-Fi Module Radio Signal ESP8266, Intel AX200 1-10 Hz Low Infrared Sensor Thermal Imaging FLIR Lepton IR Camera 1-60 Hz Medium II System Architecture & Use Cases In this section,

Read the complete text