teleoperation latency measurement

This study addresses a critical gap in the evaluation of teleoperation systems, where existing approaches predominantly focus on glass-to-glass (G2G) latency while neglecting motion-to-motion (M2M) latency—the delay from control input to vehicle actuation—thus failing to fully characterize end-to-end (E2E) performance. To overcome this limitation, the work proposes a novel E2E measurement framework that integrates both G2G and M2M latency through a synchronized assessment architecture. Using two GPS-synchronized Raspberry Pi 5 units equipped with gyroscopes, phototransistors, and LED-triggered interrupts, the system employs low-pass filtering and threshold detection to precisely identify steering actions and align events across endpoints. Experimental results over commercial 4G/5G networks reveal an average E2E latency of approximately 500 ms (±4 ms), with M2M latency accounting for up to 60% of the total, thereby demonstrating the method’s high precision and validity.

This paper addresses the challenge of evaluating control latency in remote operation of connected and automated vehicles (CAVs). We propose the first end-to-end, architecture-agnostic Motion-to-Motion (M2M) latency measurement framework, which formally defines and quantifies the delay between a remote operator’s steering input and the vehicle’s corresponding steering execution. Our method employs Hall-effect sensors and a dual-synchronized Raspberry Pi 5 system, leveraging interrupt-driven timestamping to achieve high-precision synchronization at both human and vehicle ends, with measurement accuracy of 10–15 ms. Moving beyond conventional video-link-centric latency analysis, our framework reveals—through real-world testing—that actuator response dominates M2M latency, contributing a median delay of 750 ms. This work establishes a reproducible, standardized benchmark tool for rigorously assessing real-time performance in remote driving systems.

This study addresses the critical challenge of deploying remote driving, which hinges on wireless networks delivering low and predictable end-to-end latency. Leveraging MASA—a city-scale real-world testbed—this work presents the first systematic comparison between ITS-G5 and 5G in an operational intelligent transportation infrastructure, evaluating uplink video streaming and downlink control command latency and reliability under varying network loads and traffic conditions. The findings reveal fundamental differences in latency and its variability between the two technologies, quantify the impact of infrastructure coverage on video transmission quality, and inform the design of a hybrid communication strategy. This approach significantly enhances the feasibility and safety of remote driving, offering actionable guidance for network selection and integration in future deployments.

The impact of network latency and video quality on operator workload and driving performance in remote driving remains poorly understood, limiting system reliability. This study investigates the effects of systematic variations in control latency (100/300 ms) and video bitrate (500/2000 kbit/s) through a driving simulation experiment, integrating multimodal physiological measures—including eye-tracking, electrocardiography, and electrodermal activity—to assess cognitive load and driving performance. Results indicate that a 300 ms latency combined with a 2000 kbit/s bitrate achieves speed performance equivalent to ideal conditions. Furthermore, physiological metrics reveal sub-additive interaction effects, supporting the development of a physiology-based, proactive overload warning mechanism. These findings provide both theoretical grounding and a technical pathway for optimizing remote driving system design.

This study addresses the poorly understood impact of network latency on closed-loop stability in vision-based teleoperation, particularly within perception-driven control frameworks. The authors develop a Latency-Aware Visual Teleoperation (LAVT) testbed built on ROS 2, enabling precise injection of controllable one-way delays in simulation to systematically evaluate the nonlinear effects of perceptual latency on lane-keeping performance across diverse road scenarios. Through 180 closed-loop experiments, they uncover—for the first time—a sharp stability collapse within the 150–225 ms perceptual delay range: when one-way delay exceeds 150 ms, task success rate plummets from 100% to below 50%, primarily due to phase-lag-induced oscillatory instability. The work further quantifies how additional control-channel latency accelerates system failure and establishes a reproducible benchmark for latency-induced degradation in teleoperated driving.

This study addresses the stringent requirements of remote teleoperated driving, which demands ultra-reliable and low-latency communication that current networks struggle to fulfill. The work systematically evaluates the feasibility of 5G networks in this context, with a focus on how time-division duplex (TDD) frame structures and uplink/downlink bandwidth configurations affect end-to-end latency for video transmission and control commands. It further models system scalability under multi-vehicle concurrency. The findings reveal that specific 5G configurations can support remote driving, albeit at the cost of reduced video bitrates to accommodate more vehicles. Moreover, internet backhaul latency imposes a significant bottleneck on centralized architectures, highlighting their inherent limitations in meeting the most demanding latency constraints.

This study addresses the limitations of current 5G networks—primarily optimized for mobile broadband—in supporting large-scale remote driving, which demands high uplink bandwidth, ultra-low latency, and high reliability. The work presents the first quantitative analysis of uplink bandwidth requirements for concurrent remote driving under different 5G architectures, systematically evaluating the performance gap between edge computing (MEC)-based and centralized deployments. It further investigates the impact of duplex modes, TDD frame structures, and control channel configurations on network capacity. Findings reveal that MEC significantly enhances system scalability and that optimizing control channel allocation effectively mitigates video processing latency bottlenecks. Results demonstrate that a MEC-enabled 5G architecture is better suited for remote driving scenarios, with judicious parameter configuration substantially increasing the number of concurrently supported vehicles, thereby offering theoretical grounding and practical design guidelines for 5G-based teleoperation deployment.

This work proposes a cross-layer, interpretable performance diagnosis method to address the challenge of detecting subtle radio-layer dynamic anomalies in O-RAN systems when end-to-end latency appears stable. Leveraging real-world measurements across multiple distances and user equipment (UE) types, the approach jointly analyzes application-layer tail latency—such as the 95th percentile—with radio-layer metrics including scheduling behavior, modulation and coding scheme (MCS), block error rate (BLER), and signal quality to construct lightweight “degradation flags.” The method enables non-intrusive yet effective detection of radio-layer performance degradation, revealing the sensitivity of tail latency to UE type, distance, and network load. This facilitates practical and efficient fault localization and monitoring in O-RAN deployments.

This study addresses the absence of a reproducible benchmark for evaluating remote surgical video streaming under realistic network impairments. The authors propose the first standardized evaluation framework that jointly models network conditions, video quality, and surgical task performance. Leveraging Linux Traffic Control with NetEm and a Gilbert-Elliott packet loss model, the framework simulates five representative network scenarios while simultaneously measuring QoS metrics, objective video quality (PSNR, SSIM, VMAF), temporal continuity (freeze ratio), and human operator performance. Across 375 experimental trials, network degradation caused task success rates to plummet from 97% to 12% and completion times to increase from 80 to 255 seconds, thereby providing the first systematic evidence of the substantial impact of forward video link impairments on remote surgical performance.

In the era of sub-millisecond networking, host-side latencies—such as those introduced by the kernel network stack and application scheduling—have become the dominant bottleneck for end-to-end low-latency performance, yet production environments lack effective means for continuous monitoring. This work proposes and implements netstacklat, the first system to enable low-overhead, continuous end-to-end latency monitoring within the Linux kernel network stack. By leveraging lightweight kernel probes and an efficient performance monitoring framework, netstacklat accurately captures the data path latency from the network interface card to the application across 144 diverse Nginx/Apache HTTP workloads, incurring less than 6% overhead even at tail latencies. The tool has been successfully deployed across Cloudflare’s global CDN infrastructure, demonstrating its scalability and practical utility in real-world production settings.