A virtual slotframe technique for reliable multi-hop IEEE 802.15.4e time-slotted channel hopping network

Abstract

Time-slotted channel hopping is one of the medium access control modes defined in IEEE 802.15.4e. Although time-slotted channel hopping provides high reliability, it cannot be achieved automatically. A time-slotted channel hopping should be configured properly according to the dynamic changes in a network. When new devices participate in the network or the data traffic is increased, link allocation may not be possible due to the fixed slotframe length. The simplest way to acquire additional links is to change the length. However, the conventional method to change this length involves significant overhead and the possibility of link failures. In this article, we evaluate the performance of the conventional IEEE 802.15.4e method and analyze the problems that can occur when changing the slotframe length. To resolve these problems, we propose a virtual slotframe technique that forms a logical slotframe by connecting multiple slotframes. A slot scheduler will then perceive the virtual slotframe as merely a long slotframe. The devices can translate the schedules made for the longer slotframe into real links using the virtual slotframe technique. These features allow the time-slotted channel hopping network to allocate additional slots without reconfiguration. The simulation results show that the proposed technique is a maximum of 18 times and an average of 10 times faster than the conventional method.

Keywords

IEEE 802.15.4e time-slotted channel hopping virtual slotframe Internet of Things industrial wireless sensor network

Introduction

In the vision of the Internet of Things (IoT), all of the things around us are connected to the Internet and cooperate with each other to achieve a common goal without human intervention.¹ IoT is defined in many forms and contains many different research areas. Atzori et al.² categorize the IoT in terms of three distinct perspectives: Internet, Things, and Semantics. The implementation of the IoT requires web and Internet technologies (an Internet-oriented vision to connect things to the Internet), identification and wireless sensor network (WSN) technologies (a Things-oriented vision to connect things to each other), and context awareness technology (a Semantic-oriented vision to provide intelligent services).

In this article, we focus on the Things-oriented vision. Billions of intelligent communicating things are expected to participate in the network, and the Things-oriented vision contains communication technologies to facilitate this. Heterogeneous communication technologies are expected to be used to connect things to the Internet, and low-power wireless communication technologies (i.e. WSNs and radio-frequency identification (RFID)) are considered to be most suitable to facilitate operation for several years without human intervention. WSNs have been studied for a relatively long time, and many WSN technologies such as ZigBee, Z-wave, Bluetooth, and 6LoWPAN are evolving to become essential elements of the IoT.³

Time-slotted channel hopping (TSCH) is one of the medium access control (MAC) behavior modes defined in the IEEE 802.15.4e MAC amendment.⁴ It was designed mainly for industrial wireless networks and has become the de facto standard for industrial WSNs. Industrial wireless network technologies require the highest levels of reliability and stability, and the network must be able to operate in very harsh environments. TSCH was designed for industrial applications, and these features of TSCH make it well suited for IoT environments in coexistence with various wireless communication technologies.

Although TSCH provides high reliability, this reliability cannot be achieved automatically. First, TSCH based on IEEE 802.15.4e has several network parameters, and a TSCH network should be configured properly according to the network environment. Second, TSCH is a time-slotted access technique based on multi-channel concept, and slot scheduling is an essential element to provide reliability, although the 802.15.4e MAC amendment does not specify how to make such a schedule.

The Internet Engineering Task Force (IETF) 6TiSCH Working Group (WG) proposed a sub-layer (i.e. the 6top layer) to allow TSCH to be used easily and to utilize industrial-like reliability with IoT networks,^2,5 and there is some slot scheduling research for TSCH in the IoT environment. Palattella et al.⁶ and Jin et al.⁷ proposed a central slot scheduling algorithm, and Accettura et al.,⁸ Soua et al.,⁹ and Choi and Chung¹⁰ proposed a distributed slot scheduling algorithm for TSCH. However, the dynamic changes of a network can cause a lack of network resources. For example, when new devices participate in the network or the data traffic is increased, the network resources may be insufficient due to the fixed length of the TSCH slotframe. The simplest way to acquire additional resources is to extend the slotframe length. However, the conventional reconfiguration method to extend this length using an Enhanced Beacon (EB) creates significant overhead. The conventional method adversely affects the reliability of the network for IoT services, and a new idea is needed to solve the problem.

In this article, we introduce the IEEE 802.15.4e TSCH MAC behavior mode briefly in section “TSCH,” and in section “Changing the TSCH slotframe length” we analyze the problems that can occur when changing the TSCH slotframe length. To resolve these problems, we propose a virtual slotframe technique in section “Virtual slotframe.” In section “Performance evaluation,” we evaluate the performance of the conventional reconfiguration method and the proposed idea. Finally, in section “Conclusion and future work,” we draw conclusions and suggest future work.

TSCH

TSCH overview

Industrial wireless network technologies require the highest levels of reliability and stability, and the network must be able to operate in very harsh environments. TSCH was designed for industrial applications and combines time-slotted access and channel hopping features to provide network reliability and stability. Channel hopping is a key feature of TSCH that provides agility and robustness against external interference. Owing to these features, TSCH has become the de facto standard for industrial WSNs and was adopted by WirelessHART, ISA100.11a, and as the IEEE 802.15.4e MAC amendment.¹¹

TSCH was designed for industrial applications to have high reliability and to be robust against interference. Such features make it well suited for IoT environments, in coexistence with various wireless communication technologies. The IETF WGs have been developing a Low-power Lossy Network (LLN) stack¹² to realize the IoT, and they expect that the TSCH network could open up various services to those applications that cannot currently utilize IoT wireless networks due to reliability problems (e.g. home security, smart city, assisted driving, tracking, etc.).¹³

As a part of such efforts, the IETF 6TiSCH WG proposed a sub-layer for TSCH.¹⁴ IEEE 802.15.4e only defines the TSCH mechanism for communication and does not specify a logical link control (LLC; e.g. a network policy and slot scheduling) to utilize such a mechanism. The IETF 6TiSCH WG defines an LLC layer to bridge such a gap and to control the logical link easily. Moreover, IEEE 802.15.4e defines only the MAC layer, and it is possible to utilize other IETF IoT network stacks such as Routing Protocol for Low-power and Lossy networks (RPL) and Constrained Application Protocol (CoAP).¹⁵ These efforts will allow TSCH designed for industrial wireless networks to be used widely in IoT environments.

TSCH mechanism

A TSCH network requires time synchronization, and all devices in the network must be synchronized with a coordinator. The time synchronization algorithm used for TSCH should support a multi-hop network, which means that the algorithm must handle problems such as clock drift, which can occur in multi-hop networks. Synchronization performance is a very important issue in a TSCH network, because it is directly related to the energy efficiency and reliability of the network.

As shown in Figure 1, the slotframe of a TSCH network repeats over time, and one slotframe consists of multiple timeslots. Figure 1 shows an example of a four-length slotframe that consists of four timeslots; the actual length of a slotframe depends on the network configuration. A single communication is performed at one timeslot, and the length of each timeslot is configured to allow the exchange of data and an acknowledgment (ACK). A slot offset refers to the relative location from the start of a slotframe, and it is recounted at the start of each slotframe. In contrast, the absolute slot number (ASN) refers to the order of the timeslots counted from the start of the network.

Figure 1.

Frame structure of a TSCH network.

The IEEE 802.15.4 PHY standard defines 16 channels at the 2.4-GHz frequency band. TSCH can use the channels selectively, and the number of channels used depends on the network configuration. A channel offset means an index of the available channel list, and the real frequency of the channel is calculated by equation (1)

f = F [(A S N + c h a n n e l O f f s e t) \mod N_{c}]

(1)

where N_c is the number of available channels and F[·] is a look-up table of the available channels. The slot scheduling method of TSCH relies on selecting a time (i.e. a slot offset) and a frequency (i.e. a channel offset) at which communication is performed. A link is defined as the combination of the slot offset and channel offset. If a link X (e.g. slot offset 1 and channel offset 5) is allocated between devices A and B, the two devices will use channel offset 5 for every communication. However, the ASN is changing continually and the real frequency will change at each communication according to equation (1). TSCH switches the frequency at each timeslot using multi-channel, which is referred to as channel hopping. If the state of a certain channel becomes bad, the TSCH network can use the blacklisting technique to remove the interfering frequency from the list of available channels, which in turn provides better network reliability.

Link scheduling for TSCH

Figure 2 shows a generic example of a multi-hop wireless network. If a transmission from device 3 reaches device 4, the reception of device 4 will be disturbed. However, if device 4 uses another channel, the transmission of device 3 does not affect device 4. Although the multi-channel technique allows collisions to be avoided, a single device can use only one channel because low-power wireless networking devices generally have only one transceiver. In other words, devices 3 and 4 cannot transmit to device 1 at the same time, even if they use separate channels.

Figure 2.

Tree topology of a multi-hop wireless network.

Slot scheduling is required to communicate in a stable manner without collision. Moreover, utilizing parallelism in multi-channel requires a coordinated channel selection that considers the interference relationships. A goal of slot scheduling is to ensure contention-free communication by scheduling communications that could cause a collision at different times or on different channels. For this purpose, information is required about the network, in order to establish the collision and interference relationships. Slot scheduling based on the network environment must be rescheduled according to changes in environment to provide optimal performance.

In the literature,^6–10 relatively recent slot scheduling algorithms for a TSCH network were proposed; they allocate links according to the level of traffic in a convergecast network and aim to transfer all data to the root device within a single slotframe. However, the traffic level can change according to changes in the network environment, such as the participation of new devices or a change of topology. In this case, rescheduling must be performed to comply with the basic settings of the network (i.e. duty cycle) or the goal of the algorithm (i.e. to transfer all data to the root within a single slotframe). However, restricted resources within a single slotframe could make this impossible. To solve such a problem, a slotframe management mechanism is required that can allocate network resources effectively.

Changing the TSCH slotframe length

Figure 3 shows an example of slot scheduling for devices 1, 4, and 7, which are shown in Figure 2; each device has its own links to communicate with its neighbors. In this example, if a new device participates in the network as a child of device 4, an additional link cannot be allocated to the new device due to a lack of resources. It might appear that the slot schedule of device 4 in Figure 3(b) has spare links, but these other links are not available because a device can listen to only one channel at a time. In order to solve this problem, a basic solution is to increase the length of the slotframe. However, in order to implement such a solution, EB messages should be transferred securely to each device in the network, which causes significant overheads.

Figure 3.

An example of slot scheduling for devices 1, 4, and 7: (a) slot schedule of device 1, (b) slot schedule of device 4, and (c) slot schedule of device 7.

A TSCH network advertises the EB messages in order to maintain the network; these messages contain channel hopping, timeslot, initial link, and slotframe information. Coordinator devices in the TSCH network should advertise the EB messages, and a device wishing to join the network scans the EB messages. After receiving an EB message, the device synchronizes with the network and sets the network configuration using the information contained in the EB. The coordinators have a duty to advertise EB messages to support the aforementioned sequences and announce the presence of the network.

Going back to the previous problem, if the length of the slotframe is required to change according to the network environment, the coordinators would advertise the EB messages containing the update information to the TSCH slotframe and Link Information Element (IE). Devices that receive the EB message will change network configurations, so if the configuration associated with the length of slotframe is changed, it may cause a problem.

Figure 4 shows an example of such a problem. At time t0, devices D1 and D3 are synchronized with each other and operate with a four-length slotframe, and at time t1 device D1 changes the length of its slotframe and transmits an EB message containing this information. If device D3 fails to receive the EB message or does not apply the changed configuration immediately, the slotframe of device D3 will drift and all communication after time t2 will fail. Since the slotframe of device D1 starts at time t3, but the device D3, which has not received the changed information, still starts at time t2.

Figure 4.

Example of a problem that can occur when changing the length of a slotframe (Late Change).

As shown in Figure 5, if device D3 receives the EB message and changes the length of its slotframe immediately, the start of the slotframes would be synchronized at time t3, and devices D1 and D3 can communicate successfully with each other. However, D5 (the child device of D3) will fail in every communication after time t2, because D5 is still operating using a four-length slotframe. Moreover, the misaligned schedule of D5 acts as an interference factor to the neighbor devices, and it will reduce overall network performance. In other words, regardless of the timing of applying the configuration, slotframe drift cannot be avoided. If the slotframe drift lasts for a relatively long time, a device could be disconnected from the network and lose synchronization. One solution is to synchronize the timing of applying the changed configuration by considering how much time it would take for an EB message to propagate throughout the network. However, it is very difficult to guarantee secure propagation of EB messages without loss. Even if the network can guarantee such secure propagation, calculating the propagation delay considering the loss of EB requires too complicated a process and arithmetic for constrained devices. This EB-based method forces too heavy an overhead on the devices.

Figure 5.

Example of a problem that can occur when changing the length of slotframe (Immediate Change).

Consequently, an IEEE 802.15.4e TSCH network that does not use any special techniques should change its network configuration gradually, even if a severance of the network may occur (for convenience, let the basic method for changing the length of a slotframe be called the gradual-changing method). In an IoT service that requires high reliability, the network severance time of the gradual-changing method is an important element and the method requires analysis.

Virtual slotframe

Virtual slotframe method

The basic idea of the proposed technique, called a virtual slotframe, is to make a logical slotframe by connecting multiple slotframes without reconfiguring the network. A slot scheduler will perceive the virtual slotframe as merely a long slotframe and can schedule it using more resources than it actually has (let the schedule that uses the virtual slotframe be called a virtual schedule). The devices that participate in the network can then translate the virtual schedule into a real link. Unlike the gradual-changing method, a virtual slotframe method does not cause network severance time and can provide reliability and stability, because it does not change the real length of the slotframe.

In order to describe the proposed technique, we define certain variables shown in Table 1. The virtual slotframe coefficient (VSC) is a key variable of the idea, and it decides the required number of slotframes to make a virtual slotframe. For instance, if the VSC were 3, the virtual slotframe would consist of three real slotframes. And the slot scheduler can then build a schedule using a slotframe that is three times longer than the real slotframe. For instance, in Figure 3, the slot scheduler can select a slot offset (SO_R) between 0 and 3, but it is possible to select a virtual slot offset (SO_V) between 0 and 11 when a virtual slotframe is used with VSC = 3.

Table 1.

Variables for the virtual slotframe.

Variable	Description
L_S	The length of the real slotframe
ASN	Absolute slot number—the slot number counted from the start of the network
VSC	Virtual slotframe coefficient—the number of slotframes used to construct the virtual slotframe
SO_R	The slot offset of the real slotframe
SO_V	Virtual slot offset—the slot offset of the virtual slotframe
VSO	Virtual slotframe offset—the order of the real slotframe within the virtual slotframe

The VSC is the information required to configure the slotframe, such as the length of the slotframe. The slotframe information should be shared by all devices in the network, and IEEE 802.15.4e defines the TSCH slotframe and link IE, which can optionally be included in an EB, for this purpose. The IE contains a descriptor for slotframe information (i.e. slotframe descriptor), which could be extended for the VSC information. Since the VSC means a coefficient for constructing a virtual slotframe, it is possible to set a single virtual slotframe having a sufficient length using only 1 byte. When each coordinator device advertises an EB message, the EB can include the VSC information.

Fundamentally, the devices shown in the example in Figure 3 cannot communicate with their neighbors by means of the virtual slot offset (SO_V) used for virtual scheduling, because virtual slot offsets 4–11 do not exist for a real slotframe. In the example in Figure 6, when a virtual slot offset of 5 is allocated to a device, the device must be able to discern that the allocated virtual schedule (i.e. SO_V = 5) belongs to the second slotframe within the virtual slotframe, and that the real slot offset is 1. For this, we define a virtual slotframe offset (VSO), which refers to the order of the real slotframes within the virtual slotframe. In other words, the VSO is equal to 1 in the aforementioned example.

Figure 6.

Example of a virtual slotframe structure (VSC = 3).

When a virtual schedule is allocated to a device, the device can calculate a VSO using equation (2). In Figure 6, the range of SO_V is 0–11, and by taking the quotient obtained when the virtual slot offset has been divided by the length of the slotframe (L_S), we can calculate the order of the virtual schedule within the virtual slotframe

V S O = \frac{S O_{V}}{L_{S}}

(2)

In a similar manner, the real slot offset can be arrived at by the remainder value after the virtual slot offset has been divided by the length of the slotframe, as in equation (3)

S O_{R} = S O_{V} \mod L_{S}

(3)

At the slot scheduler and each device, the VSC is an additional piece of information that is provided when the virtual slotframe method is used; the virtual schedule can be translated into the real schedule using the existing information. However, in order to utilize the virtual slotframe, one more aspect has to be considered. A device that uses the virtual slotframe must be able to calculate the VSO of the current slotframe, because the virtual slotframe consists of multiple real slotframes. When a device communicates with its neighbors using the virtual schedule, the device can calculate the VSO of the virtual schedule using equation (2). However, in order to decide whether to communicate or not, the device must be able to decide whether the VSO of the current slotframe is equal to the value calculated by equation (2). The VSO of the current slotframe can be determined by equation (4)

VSO = \frac{ASN}{L_{S}} \mod VSC

(4)

The VSO must correspond to each device in the network, and the devices can decide whether or not to correspond. The ASN is shared by all devices in the network and held at a constant value by the synchronization technique. The value obtained when the ASN has been divided by the length of the slotframe indicates how many slotframe cycles have passed since the start of the network. By taking the remainder after the number of slotframe cycles has been divided by the VSC, each device in the network can arrive at the VSO equally. In other words, a device can calculate to which VSO the virtual schedule belongs using equation (2), and the device can decide whether or not the virtual schedule is performed using equation (4). Figure 7 shows the aforementioned communication decision procedure for a virtual slotframe.

Figure 7.

Communication decision procedure.

Let us summarize the operation flow of the virtual slotframe technique:

Slot scheduling—the slot scheduler perceives that the length of the slotframe is longer than it should actually be and builds a virtual schedule using the virtual slotframe structure;

Virtual schedule translation—when the virtual schedule is allocated to devices, it is translated into a real schedule suited for the real slotframe structure using equations (2) and (3);

Execution of the schedule—by comparing the results of equations (2) and (4), the devices decide when to execute the virtual schedule.

As shown in Figures 4 and 5, when changing the TSCH slotframe length, the gradual-changing method cannot avoid network severance due to the drift of the EB slot. However, the proposed method can minimize the severance time. The virtual slotframe increases in multiples of the VSC; therefore, even when the length of slotframe is changed, the mutual EB slot position does not change.

The conventional method can also achieve a similar effect if the length of the slotframe is changed in multiples. However, if the length of the slotframe is not set to a prime number, the channel cannot be used evenly,¹³ and if the length is increased by a multiple, it cannot be set to a prime number. Moreover, if the length is simply increased by a multiple, delicate adjustment of the slotframe is not possible. However, if the virtual slotframe is set to a small prime number and then increased by the VSC, the length can be finely adjusted without suffering from the problem of biased channel use.

Local virtual slotframe

For IoT applications, such as Smart Home or Smart Factory, WSNs collect data periodically and deliver commands to actuators. Generally, when devices in a network transmit data periodically, some devices located near the root suffer heavy traffic loads in relaying the received data. Conversely, devices located far from the root have relatively light traffic loads. However, a conventional TSCH network that shares the same fixed slotframe length cannot allocate resources according to the requirements of each device under asymmetric traffic conditions.

Moreover, IoT applications can have different data collection periods, and it is difficult to configure the slotframe length considering all the applications. Let us assume application A with a data collection period of 5 s and application B with a period of 1 s. If the length is set according to the period of application A, then five links for application B should be allocated within a single slotframe. On the other hand, if the length is set according to application B, a link of application A, which is used only once every 5 s, should be allocated to each slotframe. This results in a waste of memory and energy in the constrained devices. However, if we apply the virtual slotframe technique locally, we expect flexible resource allocation to be possible according to those requirements.

The proposed technique causes the slot scheduler to operate logically using additional information (i.e. the VSC). Thus, it is possible to make only a portion of the network operate with another VSC, and this feature allows the network to allocate resources according to the level required by each device. Each device determines the VSC depending on the required resources, in a centralized or distributed manner, and the VSC of each device must be advertised using an EB. Neighboring devices that receive the EB message must record the information of each neighbor’s coefficient.

In the examples shown in Figures 2 and 3, when a new device participated in the network as a child of device 4, there was a problem: device 4 could not allocate resources to the new device because device 4 did not have sufficient resources. Figure 8 shows an example of the locally applied virtual schedule to resolve such a problem. In Figure 8(b), unlike other devices (i.e. devices 1 and 7) that are operating using VSC = 1, device 4 uses VSC = 2 to secure the required resources, and the slot scheduler allocates a link to the new device 10 using the extra timeslots.

Figure 8.

Example of the locally applied virtual schedule: (a) slot schedule of device 1, (b) slot schedule of device 4 (VSC = 2), and (c) slot schedule of device 7.

However, in Figure 8(a) and (b), devices 1 and 7 still have links related with device 4 (links A and B), and the remaining links may incur unnecessary communication attempts. However, when devices 1 and 7 receive the EB of device 4, they would notice that device 4 uses VSC = 2. Through the aforementioned procedures, devices 1 and 7 determine that there is no need to perform communications on links A and B. After that, devices 1 and 7 can avoid unnecessary communication on those links.

Although the virtual slotframe technique allows a TSCH network to allocate resources flexibly, it has the same issue as IEEE 802.15.4e. The virtual slotframe technique can be categorized as a slotframe management method, and thus it does not define how the technique would be utilized. In order to utilize the virtual slotframe technique properly, a slot scheduling algorithm is required that can satisfy the following conditions:

It must allow the slotframe to expand freely according to the requirements of each device;

It must be able to handle the effects of adjacent devices caused by expansion;

It must be able to handle conflicts that can occur when different logical schedules are translated.

TSCH multiple slotframes

The TSCH multiple-slotframe technique, which is defined in IEEE 802.15.4e, allows simultaneous operation of slotframes of different sizes. This technique can be used for similar effects to the proposed technique, but it shows a clear difference in purpose and efficiency.

First, when network resources are insufficient, the multiple-slotframe technique can create a new slotframe of sufficient length without network severance. However, this approach creates multiple unique slotframes distinguished by slotframe handle, and the slot scheduler must create and distribute a new schedule for all devices associated with the modified slotframe handle. A minimum of 4 bytes of EB packet overhead occurs to contain information for the new slotframe, and if schedule information is handed over using EB, additional overhead of 5 bytes per link is required. From the viewpoint of implementation of low-power constraint system, distributing and applying different schedules to multiple slotframes increases system complexity and packet overhead.

On the other hand, the proposed technique operates on a single slotframe structure, and it can flexibly extend a slotframe to secure resources for scheduling with lower overhead. Since the proposed technique provides a virtualized view to the slot scheduler through the VSC, the slot scheduler does not need to identify the slotframe to which the schedule is to be applied. Therefore, it is easier to manage the schedule. Furthermore, it is also possible to use multiple slotframes and simultaneously apply the proposed technique for each slotframe.

Second, the multiple-slotframe technique can be used to allocate the appropriate resources for various IoT applications with different periods. However, this approach also has a problem of increasing packet overhead and complexity and there is a limit in the local allocation of resources. For example, in peer-to-peer (P2P) communication, the multiple-slotframe technique is difficult to allocate a new slotframe only to the devices on the path. On the other hand, the proposed technique can allocate additional resources to the devices on the path by adjusting the VSC locally at the scheduler level.

Consequently, in the case of low-power networks where the length of slotframe changes frequently or multiple IoT applications operate simultaneously, the multiple-slotframe and the proposed techniques can both be considered as a way to solve the problems. However, the multiple-slotframe technique requires a more complex system and large memory footprint to manage the configuration and schedule of each slotframe, and it causes a significant overhead in a constraint low-power system. We think that if we can get a similar effect on both technologies, we do not need to endure this overhead in low-power systems. On the other hand, if the network is operated by various business operators (such as commercial network systems), strict network separation may be required for billing, security, and quality of service (QoS) provisioning. The multiple-slotframe technique could be more suitable for such a case. Compared to Wi-Fi, it is similar to running multiple networks with multiple Service Set IDentifiers (SSIDs) in a single AP to provide QoS differentiating and security management for the network. In conclusion, I argue that they have obvious differences in terms of operating environment, purpose and efficiency.

Performance evaluation

When the length of the slotframe is changed in an IEEE 802.15.4e TSCH network, the gradual-changing method (i.e. the conventional basic mechanism using an EB message) cannot avoid slotframe drift. After the change has started, the slotframe drift severs a connection to a device, and this severance will last until the device receives an EB message. It is possible that the device could receive an EB message by chance. However, it has a relatively high probability of being severed, which decreases the reliability of the network.

In order to show the comparison between the conventional method and the proposed technique, we use an OpenWSN simulator¹⁶ developed in the Python language; the configuration of the simulator is shown in Table 2. Each device is placed randomly in a 300 m × 300 m space, and the network is formed using IETF RPL.¹² We use all 16 channels of IEEE 802.15.4 PHY, and the length of the slotframe is configured as a prime number to utilize the channels equally. Each device in the network operates as a full-function device (FFD) and advertises EB messages periodically. The basic advertising period is approximately 30 s, and it is diffused randomly to alleviate collisions of the EB messages.¹⁷ After all devices in the network are synchronized with the root device, the root device changes a slotframe length (to another prime number). In this simulation, we measure the average and maximum time to reconfigure the slotframe length of all the devices in the network.

Table 2.

Simulation parameters.

PHY	Frequency: 2.4 GHz Transmission power: 0 dBm IEEE 802.15.4 channel: 11–26 (16)
MAC	IEEE 802.15.4e TSCH
Routing	IETF RPL with Objective Function 0
Topology	300 m × 300 m open area Random placement

MAC: medium access control; TSCH: time-slotted channel hopping; IETF: Internet Engineering Task Force; RPL: Routing Protocol for Low-power and Lossy networks.

Figures 9 and 10 show the number of EB cycles (one EB cycle is approximately 30 s) required to reconfigure the slotframe length according to the change in the number of devices and techniques. Figure 9 presents the performance of the gradual-changing method, and the slotframe length (i.e. horizontal axis) is changed to another prime number. In the case of the proposed technique (Figure 10), it does not change the real slotframe length. Therefore, the horizontal axis is changed according to the VSC, to which the numbers in brackets refer.

Figure 9.

Simulation results of changing the slotframe length using the gradual-changing method.

Figure 10.

Simulation results of changing the slotframe length using the proposed technique.

The line graphs represent the average number of EB cycles for reconfiguration. With the gradual-changing method (15.4e), a longer slotframe length takes more time to reconfigure. A longer slotframe length produces a lower duty cycle, which increases the probability of slotframe drift. Changing the length of the longer slotframes has a higher probability of synchronization loss; therefore, the average number of EB cycles increases. However, in the case of a simulation that uses 30 devices, the reconfiguration time is shorter than the case with 20 devices, because a device in the former case has a greater chance of receiving an EB due to the number of its neighbors.¹⁷ With the proposed technique (i.e. VS), the result shows a line that is nearly flat. Because the proposed technique does not change the real slotframe length, slotframe drift does not occur. Therefore, the devices do not lose their synchronization and EB can be forwarded stably.

The bar graphs in Figures 9 and 10 represent the maximum number of EB cycles. In the case of the proposed technique, the result is approximately the same as the average. However, in the case of the gradual-changing method, the maximum value is significantly greater than the average. Generally, the maximum value is related to the device that is farthest from the root. Therefore, in the case of the conventional method, the network severance time is closely related to the distance of a device from the root, and the farthest device is most affected by the accumulated slotframe drift.

The case of VSC = 28 in Figure 10 is an anomaly. In this case, some devices could not receive an EB before desynchronization because the slotframe length was expanded too much. If a device does not receive anything for some time, it would desynchronize and find another synchronization source. However, although a device desynchronizes from the network, the severance time is significantly shorter than that with the conventional method. This problem can be solved using more accurate clock or allocating more advertisement slots. However, a scheduling policy using virtual slotframe is left for future work and is outside of the scope of this article.

With the gradual-changing method, an EB which contains a changed slotframe length is spread out gradually. Therefore, the distance from the root is a key factor of the network severance time. To clarify this relation, we measure the packet delivery ratio (PDR) of each device until all devices in the network are reconfigured. Figures 11 and 12 show the PDR of each device according to the method, slotframe length, and maximum hop count. The bar graphs represent the gradual-changing method (i.e. 15.4e), and the numbers in brackets refer to the slotframe length. The line graphs represent the proposed technique, and the numbers in brackets refer to the VSC. With the gradual-changing method in Figure 11, one-hop devices, which are neighbors of the root, are nearly impervious to reconfiguration. However, the PDR of two-hop devices decreases to nearly half, and the PDR of three-hop devices is severely depleted. Figure 12 shows the PDR of each device in the network with a maximum of five hops, and the overall result is very similar to the former case, although the maximum number of hops is changed. As a result, in the results of Figures 11 and 12, this trend is shown regardless of the slotframe length and the depth of the networks. With the gradual-changing method, the PDR and network severance time of a device are linearly proportional to the number of hops in the network.

Figure 11.

PDR of each device during simulation (maximum three hops).

Figure 12.

PDR of each device during simulation (maximum five hops).

In contrast, the proposed technique shows a higher PDR than the conventional method. Although the results in Figures 11 and 12 show that the PDR of the proposed technique is about 90%, the number of actual losses is two or three, because the network severance time of the proposed technique is significantly shorter than that of the conventional method. The number of actual losses per unit time is 20 times smaller than that of the conventional method.

Conclusion and future work

The unstable link of wireless networks and the resource requirement of various different IoT services create a dynamic network environment, and such a situation demands a changeable slotframe length. However, changing the length of the slotframe using an EB message, which is a basic network configuration mechanism of IEEE 802.15.4e, causes significant overhead and decreases the reliability of the network. We analyzed the problems that can occur when changing the slotframe length using a conventional EB message, and we simulated the situation regarding network severance. Furthermore, we proposed a virtual slotframe technique that makes a logical virtual slotframe by connecting multiple real slotframes. The virtual slotframe technique can decrease the network severance time drastically in comparison with the conventional gradual-changing method. The simulation results show that the proposed technique is a maximum of 18 times and an average of 10 times faster than the conventional method.

IoT applications can have different data collection periods and various data traffic patterns. Some applications may report sensor data periodically, or only when an event occurs. Moreover, various asymmetric traffic patterns occur such as a convergecast concentrated in the sink and a P2P relayed between the devices. Such an asymmetric traffic load is difficult to handle flexibly with a fixed-length TSCH, but the proposed technique can appropriately allocate network resources according to the level required by each device. This feature can help each device in the network to maintain a proper duty cycle, and we expect that it could also prolong the overall lifetime of the network. However, in order to prove such an effect, a special slot scheduling algorithm is required that considers the features of virtual slotframe technique. As future research work, we will study such algorithms and show the advantage of flexible slot allocation that uses the virtual slotframe.

Footnotes

Handling Editor: George P Efthymoglou

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Energy Efficiency & Resources Core Technology Program of the Korea Institute of Energy Technology Evaluation and Planning (KETEP) granted financial resources from the Ministry of Trade, Industry & Energy, Republic of Korea (No. 20151110200040) and the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP) (NRF-2014R1A2A1A11053047).

References

Giusto

Iera

Morabito

et al . The Internet of Things: 20th Tyrrhenian workshop on digital communications. Berlin: Springer Science & Business Media, 2010.

Atzori

Iera

Morabito

. The Internet of Things: a survey. Comput Netw 2010; 54(15): 2787–2805.

Mainetti

Patrono

Vilei

. Evolution of wireless sensor networks towards the Internet of Things: a survey. In: Proceedings of the 19th international conference on software, telecommunications and computer networks, Split, 15–17 September 2011, pp.1–6. New York: IEEE.

802.15.4e-2012: IEEE standard for local and metropolitan area networks: part 15.4: low-rate wireless personal area networks (LR-WPANs) Amendment 1: MAC sublayer, Institute of Electrical and Electronics Engineers Std., 16 April 2012, https://ieeexplore-ieee-org-s.web.bisu.edu.cn/document/6185525/

Palattella

Watteyne

Wang

et al . On-the-fly bandwidth reservation for 6TiSCH wireless industrial networks. IEEE Sens J 2016; 16(2): 550–560.

Palattella

Accettura

Grieco

et al . On optimal scheduling in duty-cycled industrial IoT applications using IEEE802.15.4e TSCH. IEEE Sens J 2013; 13(10): 3655–3666.

Jin

Kulkarni

Wilcox

et al . A centralized scheduling algorithm for IEEE 802.15.4e TSCH based industrial low power wireless networks. In: Proceedings of the wireless communications and networking conference, Doha, Qatar, 3–6 April 2016, pp.1–6. New York: IEEE.

Accettura

Palattella

Boggia

et al . Decentralized traffic aware scheduling for multi-hop low power lossy networks in the Internet of Things. In: Proceedings of the 14th international symposium on a world of wireless, mobile and multimedia networks, Madrid, 4–7 June 2013, pp.1–6. New York: IEEE.

Soua

Minet

Livolant

. Wave: a distributed scheduling algorithm for convergecast in IEEE 802.15.4e TSCH networks. T Emerg Telecommun T 2016; 27(4): 557–575.

10.

Choi

Chung

S-H

. Enhanced time-slotted channel hopping scheduling with quick setup time for industrial Internet of Things networks. Int J Distrib Sens N 2017; 13(6): 1–14.

11.

De Guglielmo

Anastasi

Seghetti

. From IEEE 802.15.4 to IEEE 802.15.4e: a step towards the Internet of Things. In: Gaglio

(eds) Advances onto the Internet of Things. Switzerland: Springer International Publishing, 2014, pp.135–152.

12.

Winter

. RPL: IPv6 routing protocol for low-power and lossy networks. 2012, http://tools.ietf.org/html/rfc6

13.

Watteyne

Palattella

Grieco

A. Using IEEE 802.15.4e time-slotted channel hopping (TSCH) in the Internet of Things (IoT): problem statement. RFC No. 7554, May 2015. Fremont, CA: Internet Engineering Task Force, https://www.rfc-editor.org/info/rfc7554

14.

Dujovne

Watteyne

Vilajosana

et al . 6TiSCH: deterministic IP-enabled industrial Internet (of Things). IEEE Commun Mag 2014; 52(12): 36–41.

15.

Palattella

Accettura

Vilajosana

et al . Standardized protocol stack for the Internet of (important) Things. IEEE Commun Surv Tut 2013; 15(3): 1389–1406.

16.

Watteyne

Vilajosana

Kerkez

et al . OpenWSN: a standards-based low-power wireless development environment. T Emerg Telecommun T 2012; 23(5): 480–493.

17.

Vogli

Ribezzo

Grieco

et al . Fast join and synchronization schema in the IEEE 802.15.4e MAC. In: Proceedings of the wireless communications and networking conference workshops, New Orleans, LA, 9–12 March 2015, pp.85–90. New York: IEEE.