Introduction
The rapid advancement of Industry 4.0 and the Industrial Internet has driven the transformation of traditional factory networks toward wireless, flat, and highly integrated architectures. In smart factories, diverse industrial tasks—such as robotic control, machine vision, and environmental sensing—require differentiated Quality of Service (QoS) guarantees, including deterministic latency, high bandwidth, and massive concurrent access. However, existing industrial networks, typically structured in a hierarchical pyramid with isolated Information Technology (IT) and Operational Technology (OT) layers, struggle to meet these demands.
To address these challenges, the integration of 5G and Time-Sensitive Networking (TSN) has emerged as a promising solution. While 5G offers ultra-reliable low-latency communication (URLLC), its end-to-end latency can exceed 20 ms due to processing delays in the core and transport networks, making it insufficient for ultra-low-latency industrial control applications. On the other hand, TSN, an Ethernet-based technology, provides deterministic transmission through precise clock synchronization, traffic scheduling, and redundancy mechanisms. By combining the strengths of 5G and TSN, a converged network can support heterogeneous industrial tasks with stringent QoS requirements.
This paper presents a novel Heterogeneous Traffic Shaper (HTS) designed for industrial 5G-TSN networks. The HTS efficiently manages mixed data flows—control tasks (CT), audio/video tasks (AVT), and sensing tasks (ST)—by employing differentiated scheduling mechanisms. Through OMNeT++ simulations, the proposed HTS demonstrates superior performance in ensuring deterministic transmission for critical control tasks while optimizing bandwidth utilization for non-periodic AVT and ST traffic.
Industrial 5G-TSN End-to-End Networking Architecture
System Model
The industrial 5G-TSN architecture integrates wired and wireless networks to support diverse industrial applications. The system consists of industrial gateways, switches, base stations, and various wired and wireless devices. Industrial gateways serve as the central management entity, interfacing between 5G and TSN domains. They coordinate network functions through a hybrid centralized-distributed control mechanism, where 5G Application Functions (AF) interact with TSN’s Centralized Network Configuration (CNC) module.
Industrial switches form the backbone network, connecting base stations and wired devices via TSN protocols. Industrial base stations provide 5G wireless connectivity for mobile devices such as automated guided vehicles (AGVs) and robotic arms. Unlike traditional 5G deployments, this architecture allows localized 5G service provisioning without relying on a centralized core network, reducing latency and improving reliability.
Heterogeneous Traffic Characteristics and Requirements
Industrial 5G-TSN networks must accommodate diverse traffic types with varying QoS demands:
- Control Traffic (CT): Periodic, low-latency, and deterministic, used for real-time control commands in robotic arms and motor control systems.
- Audio/Video Traffic (AVT): Non-periodic, high-bandwidth, and moderately time-sensitive, supporting machine vision and video surveillance.
- Sensing Traffic (ST): Non-periodic, variable packet sizes, and high-concurrency, used for environmental monitoring and event alerts.
To prioritize these flows, CT is assigned the highest priority, followed by AVT and ST. The HTS maps these traffic types to different queues based on their QoS requirements, ensuring isolation and efficient scheduling.
End-to-End Heterogeneous Traffic Shaping Mechanism
CT Traffic Shaping
CT tasks demand strict deterministic transmission to maintain control loop stability. The HTS allocates dedicated time slots for CT queues, enforcing a First-In-First-Out (FIFO) scheduling policy. Multiple CT queues are prioritized based on their criticality, and a guard band is introduced to prevent interference from non-periodic traffic.
The scheduling cycle (T_CycleTime) is divided into periodic (T1) and non-periodic (T2) slots. If T_CycleTime is too short, non-periodic traffic may be starved; if too long, unused time is allocated to best-effort traffic. The guard band duration is set to the maximum frame transmission time of ST traffic, ensuring no residual non-periodic traffic disrupts CT transmissions.
AVT and ST Traffic Shaping
For non-periodic AVT and ST traffic, the HTS further divides the non-periodic slot (T2) into shared (T2^s) and non-shared (T2^ns) sub-slots.
Shared Slot (T2^s) Scheduling
AVT and ST share a common queue in T2^s, regulated by a credit-based shaper (CBS). Each queue accumulates credits when idle and spends them during transmission. The credit value determines transmission eligibility:
- If credit ≥ 0, the queue can transmit, and credits decrease at a rate proportional to the port’s transmission speed.
- If credit < 0, transmission is paused, and credits replenish at a predefined rate.
This mechanism ensures fair bandwidth allocation while preventing low-priority traffic from starving high-priority flows. The credit window size and rates are dynamically adjusted to meet QoS requirements.
Non-Shared Slot (T2^ns) Scheduling
Remaining ST traffic is shaped using a token bucket mechanism, which controls transmission rates by regulating token accumulation. Each ST flow is assigned a maximum token limit (cbs) and a token refill rate (cir). A frame is transmitted only if sufficient tokens are available; otherwise, it is delayed until tokens replenish. This approach maximizes bandwidth utilization while bounding latency for non-critical traffic.
HTS Implementation
The HTS integrates these mechanisms into a unified framework:
- Priority Mapping: Incoming traffic is classified into eight priority queues based on VLAN Priority Code Point (PCP) tags.
- Periodic Slot (T1): CT queues are served in strict priority order, with gates opened exclusively for CT traffic.
- Shared Slot (T2^s): AVT and ST queues contend for transmission based on credit values, with gates opened for eligible flows.
- Non-Shared Slot (T2^ns): ST traffic is policed using token buckets, ensuring controlled transmission rates.
The HTS algorithm dynamically adjusts scheduling parameters to adapt to network conditions, ensuring deterministic latency for CT and efficient coexistence for AVT and ST.
Simulation and Performance Evaluation
Simulation Setup
The HTS was evaluated using OMNeT++ and INET 4.4, simulating a factory network with robotic controllers, cameras, sensors, and servers. Key parameters included:
- Link bandwidth: 1 Gbps
- Shared slot reserved bandwidth: 75% of total capacity
- Simulation duration: 1 second
- Traffic types: CT (PCP 7-6), AVT (PCP 5-4), ST (PCP 3-2)
The GCL cycle was set to 22 ms, with CT slots (8 ms each) and guard bands (4 ms).
Results and Analysis
- CT Traffic Performance:
• HTS and TAS both ensured deterministic latency for CT, with CT_a and CT_b experiencing stable delays of 2.17 ms and 2.28 ms, respectively. • The guard band effectively prevented non-periodic traffic interference. - AVT and ST in Shared Slots:
• AVT and ST_0 achieved average delays of 7.87 ms and 15.84 ms, demonstrating efficient coexistence. • Credit-based shaping dynamically allocated bandwidth, avoiding starvation. - ST in Non-Shared Slots:
• ST_1 and ST_2 exhibited bounded delays (13.7 ms and 16.7 ms), significantly lower than traditional TAS+CBS schemes. • Token bucket policing reduced maximum end-to-end delays by 87.29%–94.22%.
The HTS outperformed conventional shapers by combining time-triggered and event-triggered mechanisms, ensuring deterministic CT transmission while optimizing non-periodic traffic scheduling.
Conclusion
The proposed HTS addresses the challenges of mixed industrial traffic in 5G-TSN networks by integrating FIFO, credit-based, and token bucket shaping mechanisms. It guarantees deterministic latency for critical control tasks while enabling flexible coexistence for audio/video and sensing traffic. Simulation results confirm its superiority over existing approaches, making it a viable solution for smart factories. Future work will explore dynamic scheduling algorithms to further enhance adaptability in industrial environments.
DOI: 10.19734/j.issn.1001-3695.2024.08.0268
Was this helpful?
0 / 0