RT-Ranked: Towards Network Resiliency by Anticipating Demand in TSCH/RPL Communication Environments

The academy and industry have been developing new methods and technologies to provide an efficient infrastructure for interconnection among various components. The Institute of Electrical and Electronics Engineers (IEEE) and Internet Engineering Task Force (IETF) have established Working Groups (WGs) to discuss the directions of standards development. The IETF working group, denominated IPv6 over the TSCH mode of IEEE 802.15.4e (TSCH), is one such initiative. This group is motivated by the need to integrate Internet Protocol (IP) with industrial communication technologies for ingressing in the Internet of Things (IoT) world. The WG aims to create an open standard architecture using best practices and standardised specifications, allowing integration between IETF/IP and industrial communication technologies, which demand strict latency and reliability assurances for working properly.

This architecture manages and orchestrates a Time-slotted Channel Hopping (TSCH) network using the IPv6 control plane [1]. However, the IEEE 802.15.4-2015 TSCH standard does not indicate how to create and manage this schedule since each application demands different network performance resources, which is why many TSCH scheduling proposals have been developed. This situation has presented an opportunity for the scheduling to adapt and meet the specific requirements of industrial applications.

A TSCH scheduling is responsible for computing and allocating the cells to support the network traffic. The literature has classified the TSCH scheduling approaches as centralised, distributed, autonomous, and hybrid. Whereas centralised approaches use a Point Coordination Element (PCE) to effectively collect network information, compute schedules, and spread throughout the network, distributed and autonomous approaches calculate the cell demand using metrics and negotiate the cells with their neighbourhood.

For building and managing a TSCH scheduling, policies play a fundamental role. RFC 8180 specified a minimal mode for building a TSCH network [2], which defines a single shared timeslot per slotframe and allows the implementation of schedule solutions using unallocated cells. This strategy enables TSCH schedules to be built according to application requirements using unallocated cells.

Employing the previously noted minimal mode implies setting the Routing Protocol for Low power and Lossy networks (RPL) mode to non-storing, which means network nodes do not store routing information for the entire network. Instead, the routing decisions are made hop-by-hop, i.e., the complete awareness of Destination- Oriented Directed Acyclic Graph (DODAG) is only located in the root node. This RPL mode increases challenges for building the TSCH schedule; the lack of network path awareness makes it complicated for forwarding nodes to deal with traffic generated from unknown network areas.

Concerning distributed scheduling approaches, nodes use information from neighbouring nodes to implement the scheduling algorithm locally. The lack of full network awareness and its position in the DODAG, may make it difficult for forwarding nodes to handle unpredicted traffic that can originate from deep-located nodes. Some solutions use the RPL rank and the premise that the nearer the node’s sink, the higher the probability it is a forwarding node. This premise might be true if the tree generated by RPL is balanced. Still, there is no guarantee that the path tree generated will be balanced. In an industrial deployment, nodes are implanted and spread along the plant; the generated RPL path may create a branch covering dense sensor areas against others.

In this context, using a pure RPL ranking metric, which solely considers the node’s position, is inadequate to utilise provisioned cells effectively. This approach may result in allocating cells that will not be utilised, leading to unnecessary energy consumption and additional overhead during the negotiation process. Conversely, if the allocated cells are insufficient to handle traffic bursts, it may result in network performance degradation. It is important to note that scheduling and leaving cells unused increases Radio Duty Cycle (RDC) and adversely affects energy consumption. If the node knows which resources are allocated by other network levels, it can help the node’s TSCH scheduling algorithm to anticipate and compute a cell demand that can handle additional traffic efficiently.

This work proposes a solution to address potential network disruptions caused by unexpected traffic from deep-located nodes, in which we consider nodes far from the forwarding point. Specifically, supplying the schedule function with the number of allocated resources (TSCH cells) to assist it in computing a forwarding node’s schedule to support unpredicted bursts traffic. To provide this means, this proposal suggests embedding an RPL control message option that contains the number of reception cells allocated by the node. This option will be disseminated downwards via DODAG Information Object (DIO) messages and upwards via Destination Advertisement Object (DAO) messages to the network. By providing this information, the scheduling function can better predict traffic patterns and prevent issues such as disconnections, energy waste, latency fluctuations, and dropped packets, ultimately contributing to improved network stability and energy efficiency.

This paper is organised as follows: in Sect. 2, a background about TSCH, scheduling and RPL is presented. Next, in Sect. 3, the proposed ranking mechanism is presented. After that, in Sect. 4, the proposed strategy is compared to RFC 8180 (Minimal Configuration) and the Orchestra schedule. Finally, Sect. 5 concludes this paper and discusses future work.

The Industrial Internet of Things (IIoT) is a vital enabler of the fourth industrial revolution, also known as Industry 4.0, which aims to transform production processes and cyber-physical systems with the help of big data and analytics. As Goel et al. noted [3], IEEE 802.15.4 is a well-known connecting technology for IPv6 over Low-Power Wireless Personal Area Network (6LoWPANs) and comes with all the challenges that must be addressed. Using mesh topology, data is delivered using intermediate relay nodes and sensed information reaches the sink node. In this context, the TSCH and RPL protocols are commonly used to provide reliable and energy-efficient communication between nodes in Wireless Sensor Network (WSN) used by IoT systems. In this field, the TSCH and the Routing Protocol for RPL have emerged as popular protocols for enabling reliable and energy-efficient communication between nodes. Reliability is a critical requirement for many industrial applications. It ensures that messages are delivered correctly and on time. The required reliability level is typically measured by three metrics: latency (the time it takes for a message to be delivered from the source to the destination), Packet Delivery Ratio (PDR) (the ratio of the number of messages that are delivered successfully to the total number of messages that are sent), and the application threshold (the maximum number of messages that are allowed to be delivered outside of the desired time window). By ensuring these metrics are met, industrial applications can be confident that their messages will be delivered reliably, even in harsh environments. Through the use of TSCH, nodes can achieve synchronised transmission and reception of data during defined time slots. RPL delivers a scalable and robust routing mechanism for Lowpower and Lossy Network (LLN). This section will provide a background of TSCH and RPL, highlighting their key features and how they work together to enable reliable and energy-efficient communication in IoT.

A mesh network is a type of network architecture in which multiple devices, referred to as nodes, are interconnected to form a decentralised and self-configuring network. These networks rely on forwarding nodes to maintain network stability by managing traffic, handling network maintenance, and processing application messages. Mesh networks have extensive applications in industrial environments, where their distinctive features can be effectively utilised.

Nodes within a mesh network are responsible for managing the traffic of their local processes and the traffic generated by their child nodes. They are also tasked with handling network maintenance and application messages. It is important to note that in a TSCH network, the minimal operation mode defined by RFC 8180 allows a node to transmit once per slotframe, which is sufficient for bootstrapping the network. However, it is essential to acknowledge that this single slot may be inadequate to handle high transmission demand. Therefore, the TSCH scheduling function can be customised to allocate additional cells based on the specific needs of the applications involved.

When using a distributed scheduling strategy, the amount of information available is limited, making TSCH scheduling challenging. It is important to note that the minimal operation mode defined by RFC 8180 is for RPL in the non-storing mode, which limits the downward routes awareness. This adds additional complexity because, even when using RPL for deployment, there is no guarantee that the generated paths will result in a balanced tree with the sink node at the centre; from the application point of view, this may inject complexities since the network topology impacts the latency. Moreover, in a TSCH network, each node negotiates cells according to its demand, so there is no guarantee that a branch with more nodes will demand more resources than another. Balancing latency, bandwidth, and energy consumption is critical in such networks. Even though allocating more transmission cells may seem like a solution, it is essential to focus on resource management to avoid waste and shortages.

To address this issue, we propose a mechanism that enables forwarding nodes to be aware of the demand from their child nodes, allowing them to effectively manage traffic from the farthest reaches of the network. This new method, integrated into RPL control messages, requires only two bytes of payload and eliminates the need for additional message exchanges. Retro-compatibility is ensured, as specified in the RPL RFC. If the receiver does not recognise the Option Type value for an RPL option in an RPL message, it silently ignores the unrecognised option and continues processing the following option. The TSCH scheduling function can use the demand as a topological policy, independently or in combination with the node’s rank. By utilising control messages to transmit demand information, we avoid introducing new message types or protocols that could potentially increase overhead.

The proposed procedure handles incoming RPL control messages, allowing the extraction of the cell allocation information and calculation of the demand from neighbouring nodes. As exemplified in Algorithm 1, our mechanism consists of two procedures: ProcessIncomingDAO and ProcessIncomingDIO. These procedures utilise option type 0x10 to retrieve the demand value from the control messages and store it using the neighbour abstraction. The scheduling function then uses the ProcessRTDemand procedure to obtain the demanded cells in the downward and upward directions. This cell allocation awareness assists in calculating the correct number of cells, e.g., in applications where the flow from the root to the nodes is crucial (e.g., handling actuators), where the upward value becomes the most critical parameter. Conversely, the downward value is utilised in monitoring systems where the nodes primarily function as sensors transmitting information outside the LLN. The proposed approach is depicted in Algorithm 1.

RPL control messages are identified by the codes presented in Table 1. At this point, the focus is on the behaviour of DIO and DAO messages. As noted in Sect. 2.2 and pointed out in Fig. 3, DIO messages are generated by a period managed by the trickle timer and contain information about the network; DIO messages flow from the sink node to the leaf nodes. In turn, DAO messages are generated by the node to advertise destination information upward along the DODAG.

These control messages have the option field, which is used to pass information regarding the network. Table 2 shows options defined in RFC 6550 and the proposed new option in bold. The main goal is to create a new option representing the amount of transmitting cells allocated by the TSCH MAC layer. The node which receives the message should store the demand data. The data originating through DAO messages are related to the demand below the node, and data arising through DIO messages are associated with the above-direction demand.

The proposed method enables forwarding nodes to be aware of the demand from their child nodes. It can be transmitted in RPL control messages as two-byte sub-options without introducing or generating additional messages. TSCH scheduling can optimise the allocation of network resources, such as bandwidth, power, and time slots, based on information received from deep-location nodes. By utilising demand information obtained from deep-location nodes, TSCH scheduling can significantly improve the overall performance and efficiency of the wireless communication network, facilitating reliable communication and connectivity even in strenuous circumstances caused by harsh network topologies.

This section discusses our experimental evaluation of the proposed strategy we have previously presented.

This study assesses by comparing the proposed solution, implemented using the schedule presented in [19], with two existing approaches: the 6TiSCH minimal mode defined in RFC 8180 [2] and Orchestra [20]. The minimal mode serves as a baseline for benchmarking scheduler performance, while Orchestra is presently the most commonly used scheduler that autonomously allocates slots for multiple traffic planes. In the proposed method, RT-Ranked, the implemented topological rank policy is replaced by utilising information from RPL Control Messages to compute cell demand.

The experiment was designed to stress the network and reach a congestion point. A congestion point is characterised by a lack of scheduled cells to sustain application and control traffic, resulting in potential latency disturbances, high retransmission rates, and node disconnections. All nodes transmit data to the sink node in a specified time interval. The assessment is done by latency, PDR, number of RPL messages and generated messages metrics.

This work presented a novel approach for communicating the number of TSCH-allocated cells throughout the network by embedding a new option in RPL control messages. The method avoids additional control messages and piggybacks on existing ones, enabling forwarding nodes to effectively manage traffic and improve resource management regarding latency, bandwidth, and energy consumption. By representing the amount of transmitting cells allocated by the TSCH MAC layer in RPL control messages, our proposed method allows for the computation of the correct amount of cells based on the target application. Overall, this approach can potentially enhance the performance and stability of mesh networks.

By utilising this method, TSCH schedules can proactively prepare the network to handle traffic bursts from deep-location nodes, particularly in DODAGs with dense allocation branches. The ability to predict the number of allocated cells allows for implementing countermeasures to prevent traffic-related issues.

The proposed mechanism has some limitations. For example, it does not consider the traffic patterns of the network. In future work, we plan to address this limitation by developing a more sophisticated mechanism to consider the network’s traffic patterns. Overall, we believe that the proposed mechanism significantly contributes to the TSCH scheduling field. It has the potential to significantly improve the performance and reliability of TSCH schedules in challenging network environments.