Hybrid approach for evaluating dynamic scheduling mechanisms in IoT cognitive sensors for enhanced network performance

Abstract

The Internet of Things is an intelligent network of cognitive sensors that are embedded into nearby objects and appliances to create a connected environment. It collects, processes, and transmits data for automation and decision-making while providing ubiquitous services. Sensors create primarily redundant data, with several repeating bit streams and matching patterns. Due to resource restrictions, sensors are carefully scheduled to activate and deactivate sensing operations. As a result, it is highly recommended to create schemes that can handle large volumes of data while maintaining a careful strategy for data collection. This article has identified and analyzed various approaches for comprehensively scheduling cognitive sensors and then evaluated the efficiency of the three schemes, CADS, EASS, and EDASS. Based on these schemes, a hybrid approach is proposed. The hybrid scheme, “Hybrid of CADS, EASS, and EDASS” (HCEE), is recommended to be used in those areas where each scheme is implemented and evaluated for applicability individually. Using the best features of these schemes, one combined scheme can be used to provide better results than those three in their individual forms. The basic three systems were evaluated with HCEE, for communication overhead and latency, node processing and buffering, the number of messages in the network, and energy usage in various scenarios. Every scheme has its primary goal, based on specific factors and a unique set of applications. Each of them exhibits unique behavior and has a variety of uses, but using one scheme in all these scenarios has achieved better results than any of these schemes. In performance analysis, CADS and EASS are more energy efficient with 67.58 and 64.25 mean values, respectively. HCEE has an extra message overhead of 6.61% compared to the average of the three schemes. HCEE has a transmission delay of 60.07% compared to other systems, and requires less buffer than CADS but more than EASS and EDASS.

Introduction

Sensors integrated with the high processing speed of edge devices, together with the widespread connection of 5G/6G, allow the Internet of Things (IoT) to be used autonomously and conveniently in a connected world^1,2,3. Due to these advancements, devices are now more intelligent and interactive, responding to any situation efficiently. The resultant paradigm comprises device-centric systems that function autonomously with zero clicks and provide services based on context and user mood^4,5,6. These services are available for anyone, anywhere, anytime (AAA), with the control of all objects instantly. The resulting mesh network of sensors and embedded devices creates a huge volume of data. Broadcasting and managing all of these packets for resource-restricted sensors is a big challenge⁷. These sensors constitute the foundation of the IoT, which is essential to change the device-centric approach to a user-centric one by offering ubiquitous services and seamless connection based on the user’s mode and context^8,9. It brings a significant change in automated applications, including energy production and transmission, intelligent transportation systems (ITS), healthcare (e-health and telemedicine), agriculture (e-farming), smart homes, campuses, and cities^10,11,12. They are used to create smart spaces and automate information processing. Consequently, end-users benefit from a more reliable and convenient IoT system that makes their lives easier and more secure¹³.

With the increasing number of IoT devices, the underlying sensors feel more burdened with processing and communication. The underlying network that connects these devices is a cognitive sensor that is responsible for collecting and processing data on a wide array of IoT devices and appliances^14,15,16. They use IoT devices with different platforms, architectures, and protocols to connect and provide services based on the user’s mode and situation. These sensor inputs are used by IoT devices to deliver context-aware and user-centric services, modifying their behavior according to operational modes, environmental factors, and user preferences. All of this results in a multilayered service architecture that hides the underlying complex architecture from the end user with seamless service delivery from interconnected devices. However, with the hyper-connectivity of these devices, massive amounts of real-time data are produced, which have a greater impact on system performance, and it is mentioned as a big challenge in IoT infrastructure. It needs to process data in real time, without latency or delay, to make better decisions. Using sensors, IoT devices help to optimize energy consumption, lower latency, and guarantee high availability^17,18,19.

Cognitive sensors accumulate and transmit data from their surroundings to other sensors and central authorities for decision-making and further processing. The collected data have many repeating bit strings and repeating patterns. Frequent communication between these devices creates congestion and network overhead and consumes a significant amount of energy^20,21,22. Similar broadcasts have had a significant impact on other resources. The system resources are used for undesired data gathering, wasting several valuable sensing cycles. Many schemes have worked to solve the problems of optimizing resources due to excessive sensing. They have suggested many directions to improve the quality of services (QoS) in terms of energy, packet loss, delay, and other stability factors^23,24,25,26. We have thoroughly incorporated all of these schemes and analyzed the key concepts to address these issues in sensors. In the second section, we examine the three methods that are used and demonstrate improvements in energy and other characteristics.

The enhancement of energy in the IoT is mainly controlled and adjusted using dynamic voltage scaling (DVS)^27,28 and dynamic power management (DPM)^29,30. The DVS is a technique for optimizing power by adjusting the voltage and frequency according to the load. The processor uses low voltage and a lower frequency if the task requires less processing load, while it may increase with increasing processing load dynamically. The main purpose was to overcome the extensive use of energy, while less power is needed^31,32. In DPM, the system controls its internal components to control and adjust the power management internally. The system allows only those components that are needed in processing to be in an active state while keeping all others in sleep if they are not needed. It is also used in those systems that face the energy and power optimization issue, such as IoT^33,34. Figure 1 demonstrates how DVS and DPM influence energy in IoT sensors.

Fig. 1

Energy enhancement using DPM and dvs.

The investigated schemes are the three significant schemes: CADS³⁵, EASS³⁶, and EDASS³⁷, which have already been worked on and demonstrated energy efficiency and other parameters for network optimization. These schemes work by controlling the flow of bits at the sensor level and adjusting the network traffic accordingly. They use scheduling with a four-state transition model to change sensor states to switch on/off modes. CADS uses the analyzer module to check the contents of the detection packets and transmits different control messages to the leaf sensors to adjust the states. The analyzer modules at base stations (BS) or central points (CPs) are programmed and trained to check the content and adjust the sensors accordingly. However, EASS checks the content at the sensor level and eliminates the delay loops and latency that are created when broadcasting the control messages. The sensor buffer is vital in this case, and many packets need to be kept in the buffer for comparison and decision-making. In this scheme, sensors are deployed on the sensor grid for uniform resource usage and collective task management. In EDASS, the new event detection strategy has been implemented, and active sensors transmit control messages to other state sensors to become active to check the event direction and intensity. EADSS is very useful for event detection and determining the direction of these events.

Motivation

Energy efficiency and resource optimization for IoT networks are key challenges that directly impact system efficiency by prolonging network life. The three schemes (CADS, EASS, EDASS) have contributed significantly by addressing key issues such as reducing extensive sensing and communication processes at the sensor level. They also minimize network traffic, resulting in a more stable topology and enhanced energy efficiency. However, all three schemes focused on individual problems in their area of application. CADS focuses on centralized management with an analyzer module, EASS focuses on a sensor grid to localize the verification of the content of data, and EDASS mainly works with the mobility of the event and its detection. To provide a comprehensive and scalable solution, a novel hybrid scheme based on these three schemes has been proposed. It comprises CADS centralized administration for overall central management, EASS self-configuration for latency and delay reduction, and EDASS features for new event detection. The proposed hybrid mechanism improves energy efficiency, minimizes delay and latency, establishes a more stable network, and never misses an event.

Contributions

The key contributions of this review work are as follows:

• A review of the latest real-world applicability of scheduling algorithms and mechanisms, demonstrating how they can be modified to use in emerging fields such as industrial automation, smart cities, healthcare, and autonomous vehicles.

• Classification of sensor scheduling and power management techniques, based on some performance metrics with their relevant details and applicability in the real world.

• More specifically, to analyze three scheduling schemes, CADS, EASS, and EDASS, for different parameters and to determine similarities and differences between these schemes.

• Propose a hybrid mechanism (HCEE), based on the functional features of CADS, EASS, and EDASS, and check the applicability and evaluation with these schemes.

Organization of the paper

The rest of the paper is arranged as follows: “Related work in adaptive and dynamic scheduling” is ‘the related work in this review based on some parameters; “Similarities in CADS, EASS and EDASS” is the similarities between CADS, EASS and EDASS; “Evaluation and comparative analysis of CADS, EASS, EDASS, and HCEE” is the evaluation part of comparing different factors for evaluation in CADS, EASS, and EDASS; while in “Discussion”, we have included some discussion about the three schemes; and finally, the paper is concluded in “Conclusion and future directions” with some future directions.

Related work in adaptive and dynamic scheduling

The main aim of adaptive and dynamic scheduling is to create an efficient IoT system that uses optimized techniques to enhance multiple parameters. These parameters may include energy, message overhead (control messages and data packets), communication delay, processing overhead, and memory requirements. In addition to achieving these primary goals, they can also benefit from some secondary goals of a stable cluster, fewer reconfigurations, reduced network congestion, and extended network lifetime. In the following sections, all schemes are arranged based on these parameters, with a discussion of the impact on the IoT system resources. The hierarchy of all schemes that have worked on energy efficiency in IoT sensors is shown in Figure 2.

Fig. 2

Literature review for adaptive and dynamic scheduling.

Scheduling for improving energy efficiency

In this section, we present some of the latest adaptive scheduling schemes concerning energy efficiency that have been discussed. Most of the schemes in energy efficiency in sensors focus on upper layers, while here we are using the lower-level details of sensors to optimize the energy efficiency.

TSRA³⁸ is basically proposed for task scheduling and resource allocation in self-powered sensors. It uses a stochastic optimization technique to divide the large and complex problem into smaller chunks. Optimization is used in task scheduling, CPU frequency, energy utilization, and resource allocation. With extensive experiments, it has been proven that TSRA enhances energy efficiency. EASF³⁹ is an adaptive sampling mechanism for spatio-temporal correlation based on remaining energy. The two phases are learning and adaptive sampling in each cycle for the disjoint sampling sensors. All active sensors collect data and send it to the base station, where the base station analyzes the data to recover non-sampled data. Better simulation results, but scalability and other issues hinder the use in real-time scenarios. DSOM⁴⁰ is based on predictive maintenance to detect and prevent early mechanical equipment failure. The main aim was to build TPN models for the PdM for the dynamic scheduling problem of UFMSs in IoT. The user needs to utilize TPNs to express the particular problem. Machine information and domain knowledge are used to target the fault places. Mixed-integer linear programming is created and formulated using the TPN paradigm. The system is good in real-time implementation, but it is very complex and resource-starved. ESS⁴¹ has used the policy of regulating the sleep times for sensors using energy consumption and operational patterns. The main aim was to increase the lifetime and reliability of IoT devices. It schedules sleep using CNN-based pattern recognition, but is not concerned if any sensors take a longer time to wake up. TACE⁴² used automation in the building with the best interaction compromise between thermal comfort and energy savings. It compares scheduling-based, IoT-enabled control for smart buildings. This system analyzed energy, distance, and combined load and distance. EERM⁴³ used an edge-fog-cloud IoT architecture for real-time IoT applications. With better QoS, it utilized the optimal node selection. Genetic, modified genetic, and delay-aware algorithms. It balanced two factors of energy and Qos applications in an edge-fog-cloud architecture, but some parameters are skipped in the simulation, and it degrades performance with nodes increases.

Table 1 summarizes the approaches utilized for energy enhancement, whereas Table 2 shows the parameters that directly impact energy-related performance.

Table 1 Energy enhancement schemes: main ideas, advantages, and drawbacks.

Table 2 Analysis based on different factors in energy-based Schemes.

Message overhead in adaptive and dynamic scheduling schemes

There are many messages used to control and adjust the sensors in the IoT. These are $C_{Messages}$, and despite the original data, they also create an extra overhead at all collection points, such as cluster heads and base stations. The following are some of the state-of-the-art schemes.

CACS⁴⁴ uses quantum IoT sensors to reduce overhearing and idle listening. It uses the observed data packets and responds based on network conditions. It works on reducing frequent communication over communication channels by adjusting network traffic. Reduces delay using baud rates and BER, and enhances energy. It is useful in minimizing the message overhead, but due to the quantum concept, it needs more resources. NPUSCH⁴⁵ uses a hybrid optimizer mechanism to schedule, allocate resources, and adapt links. Energy enhancement utilizes the signal ratio for IoT devices. It uses round-robin scheduling with NPUSCH to schedule IoT. Improve energy efficiency and scheduling processing the RR, but signal reception and optimization are other issues that affect IoT efficiency. ICA-IoT⁴⁶ uses different factors for scheduling, including battery levels, free space loss (FSL), hop count, and energy consumption. The base is a multi-sensing optimization to evaluate different routing techniques and improve network performance. There is a balance between energy economy, latency, and dependability, but it cannot solve the fault tolerance in a congested IoT. APS-IoT⁴⁷ has proposed next-hop critical packet selection, preemptive packet queuing, multilayer priority packet classification, and contingency packet migration. It uses the migration coefficient and the ratio of per-level deadlines. Prevent data loss by applying the concept of transfer of emergency data. It also encourages equity for less urgent data and ensures the lowest waits for critical data, but message queueing is a big issue in resource-restricted environments like IoT sensors. DMRSM⁴⁸ is used in delay-sensitive multisensor scheduling with sensor queuing. It uses the distribution features of the mine tunnels to create a mixed network of wired and wireless sensors. Due to many base stations, there is an NP-hard real-time routing scheduling problem. The greedy and heuristic approaches are used as the foundation for the appropriate solution and confirm some better results, but there is a very poor level of implementation. PCDE⁴⁹ utilized the identification of a transmission graph free from interference and collisions. Within the specified slot frame size, it schedules sensors with varying data rates. It produces better results than TASA. It basically works on slot frame size, throughput, latency, time complexity, and PDR. ASUM⁵⁰ used an adaptive scheduling technique with MQTT to optimize IoT communication. Adjusts transmission intervals in response to network congestion for parameters such as temperature, humidity, and pressure. It implements a mathematical framework that adapts adaptive scheduling to increased throughput, reduces latency, and improves energy efficiency. Better in memory management, but affects communication by message delay. DMRS⁵¹ is a multi-base-station routing scheduling technique. It uses irregular distribution features of several tunnels in the mine construction environment. A multi-sensor and multi-base-station real-time scheduling problem-based architecture. It uses heuristic and greedy strategies for appropriate solving algorithms. Better for memory management for heavy processing and utilizes more energy. All the schemes in adaptive and dynamic scheduling have been summarized in Table 3, and the main parameters that are affected by message overhead are mentioned in Table 4.

Table 3 Schemes with message overhead in adaptive and dynamic scheduling.

Table 4 Impact of key factors on message overhead for each scheme.

Comparative delay evaluation

Delay is an important factor in IoT sensors because the timely arrival of data is very crucial for real-time applications. The following are some of the state-of-the-art schemes that have worked on delay with other factors.

MADT-IoT⁵² is based on a Markov decision process with a Lagrange multiplier to loosen the hard constraint of channel resources. It divides it into optimization subproblems and multinode scheduling for the master policy. It provides a base for Lyapunov and linear programming approaches. Deep reinforcement learning has been applied to scheduling opportunities and slot allocation. According to numerical studies, the suggested strategy can meet the delay limitation, but it needs more sophisticated optimization methods. WISE⁵³ is based on a time-sensitive traffic scheduler with agent-based deep reinforcement learning. IEEE 802.11 networks for exclusive channel access provide a basis for the proposed TSN. This framework utilizes the latency caused by the shifting wireless settings and understands the repetitive transmission patterns. It ensures better results in obtaining scalability, but it covers only delay; many other factors can affect system performance due using the same parameters. DSPA⁵⁴ is the EFC continuum’s dynamically available resource capacity and the fluctuation-based mechanism. Various resources and operators are used in composing the DSPA application. They use assessment response time and resource use costs, which change the EFC continuum. The response time and EFC resource capacity limits also define the adaptive scheduling to minimize the cost of using the shared resources. SPEA-II⁵⁵ is used to reduce energy and system response time by taking the load balancing and deadline constraints. It is the modified version of SPEA-I, with some changed operators to obtain the best scheduling mechanism for NP-hard problems. Performing better in terms of energy and delay, but resource starvation can lead to increased energy consumption. IPAQ⁵⁶ is a time-aware scheduling algorithm that divides tasks based on their priority over time. The low-time sensitivity tasks benefit from faster reaction times, while the high-time sensitivity tasks are given priority. It also combines the AHP with an improved particle swarm optimization to discover the best task scheduling. E2rC⁵⁷ is a service migration technique to optimize the configuration of computational resources. It uses deep reinforcement with a Markov multi-phase decision system. For better service management, it uses fewer end-to-end delays and uses less energy. It maintains lower energy usage by minimizing delay, but it still requires resources. ISL-ITMH⁵⁸ is used for energy optimization and trust management in IoT. It uses energy efficiency, packet delivery ratio, routing overhead, and end-to-end delay to improve the core model. It enhances throughput and data delivery rate, but in some cases, it needs a trade-off; otherwise, it degrades system performance. In Table 5, the Summary of Delay-Minimizing, while in Table 6, all the factors are mentioned in adaptive and dynamic scheduling.

Table 5 Summary of delay-minimizing scheduling schemes in adaptive and dynamic scheduling.

Table 6 Factors affecting delay in adaptive IoT scheduling.

Processing complexity and memory analysis

These two factors highlighted the internal functionality of the sensors. Energy consumption increases with massive processing complexity. It also affects the time lapse when collecting data by limiting the real-time response time. The following are some of the schemes used in the IoT system to handle processing complexity and memory consumption.

CCIA⁵⁹ is a wavelet-based edge computing created for low-power IIoT sensors. Improves the wavelet transform by using lightweight instructions without compromising data quality. It compared the assembly language and C-based implementations for testing the industrial vibration dataset using an ARM Cortex-M7. Although better results are achieved, many technical details are still assumed in implementation, which may affect the system’s performance. Radar-PIM⁶⁰ is an embedded processor framework for Radar-PIM architecture. It provides an optimization technique for Radar-PIM processors, which is highly recommended for less energy and memory. Its performance is more than three times that of the multicore processor with the same power consumption. D2D-IoT⁶¹ is based on CoAP, MQTT, and MQTT for Sensor IoT. Its primary function is to assess possible compatibility and interoperability issues. It also works on the protocol’s scalability and capability for managing a large number of IoT devices. Implemented in ProtoLab and got better results, but the network emulator with client and server has many practical challenges that were not considered, and it skips many details. CACIE⁶² is based on DFH, which is a trustworthy metaheuristic technique for the fitness function. The optimization function is used in proximity, distance, remaining energy, ASBO, PDFH rotation frequency, and the degree of data fusion group (DFG). Improved the function for throughput, average latency, energy consumption, and network longevity. DD-HAR⁶³ is a distributed intelligence and dynamic HAR architecture. It uses real-time predictions, stored outputs, and trained models. For location accuracy, wearable sensors and cellphones, such as accelerometers and gyroscopes, are used. This system can adapt to any change and can be trained with the available and previous data. PIM-IoT⁶⁴ uses PIM topologies to minimize energy consumption. It utilizes a hierarchical system with three tiers-sensing. Jobs are combined in tiers and examine the data flow, and apply a collaborative PIM structure for each layer to efficiently manage the data processing. Although it is a pioneering work in this area, it is an optimization of a multi-tier system under power and latency limitations. MA-LHTO⁶⁵ is a multi-stage mixed-integer nonlinear programming model. It is used to maximize GPU fragmentation rate and system processing capabilities. It is coupled with a deviation-based Lyapunov optimization framework that successfully balances resource consumption and system stability. It maintains the memory use around a predetermined optimal threshold. It follows the MA-LHTO algorithm, a multi-agent deep reinforcement learning approach. All these schemes are summarized in Table 7, and in Table 8, the summary of all parameters is shown, which are affected by processing Complexity and Memory Analysis.

Table 7 Summary of processing complexity and memory analysis in adaptive and dynamic scheduling.

Table 8 Parameters affecting the processing complexity and memory analysis.

Similarities in CADS, EASS and EDASS

The main objective of these three schemes is to schedule the sensors according to their functions and according to the content of the data. Although the content analysis policy has remained the same, they are implemented in different places, and each of them has followed a different procedure. The HCEE is the uniform way of content analysis at the sensor level, and then transmits some control messages. In the following sections, some of the similarities are listed.

Data content analysis mechanism

These schemes analyzed the contents of the detected data packets and performed different actions appropriately. Sensors continuously sense the intensity of an event and convert these analog values to a digital format. They also share these values with their neighbors’ sensors and send them to CPs/BS for further analysis. They check data on bits-by-bits, pattern matching, or frequency distribution. In these techniques, the content check of detected data packets was employed to control sensing capabilities and reduce frequent communication. Typically, a sensor repeatedly senses an event and transmits the packets to the CPs/BS. In such a huge volume of data, the same data are redundantly marked and sensed again and again, which not only wastes resources but also creates congestion in the system^66,67. By checking the contents of data packets, the sensors are adjusted in different states as necessary to reduce excessive sensing in the sensing module, which minimizes the over-cost processing, and all these will lessen the frequent of communication on radio links. Dropping redundant or unnecessary data from the network improves system efficiency while lowering network traffic and congestion.

In the first scheme (CADS), the contents of the data are analyzed and tested for similarities at the analyzer unit at BS/CPs. The Analyzer unit is an intelligent and resource-rich section of BS/CPs for content checking and testing. Sensors get data from the environment and send it to BS/CPs. The Analyzer analyzes the contents of data for repeated bit strings or data patterns. If similarities are found in data repeatedly, the analyzer sends different $C_{Messages}$ to those sensors who involved in redundant sensing. Now, receiving these $C_{Messages}$, the sensors adjust their current state to a new state accordingly. In CADS, the analyzer units and $C_{Messages}$ are essential components and are required for the Four State Model to be implemented.

In the second scheme (EASS), data is collected by sensors, but the contents are checked at the sensor level by using the self-configuration technique. The sensor itself works like a content-checking entity, while the sensors are capable of automatically changing or switching states. In the first step, sensors collect data and keep a copy of the same data or a signature of these packets in the buffer. In the second step, new data is collected; it compares these values with buffered values and sends them to the BS. In the last step, if similarities are found, the sensors can change their current state to another state. This process is repeated, and the sensor state is altered accordingly.

EDASS analyzes the contents of detected data packets and broadcasts a message to all sleeping sensors to become active. The sensors are deployed in an active state, and after certain sensing cycles, they are scheduled in different states. The total sensors adhere to the Four State Model to ensure multiple states and a dynamic sensor topology are established. When an active-state sensor detects an updated event, it broadcasts an awakening message to all neighboring sensors, prompting them to become active from the other three states. It activates its states to detect the same event after receiving the message. In EDASS, as with other schemes, the content checking is still important, and any decision is made on this content inside the detected packets.

The hybrid of these three is the HCEE, which works on checking the content of the data packet at the sensor level and needs a buffer for comparison. It analyzes the content sense data and sends a message to all sleep-state sensors to activate for the detection of the new event. The sensors are placed in an active state and are configured to change states after a particular number of sensing cycles. It works like EDASS, while other functions are different. The sensors in EASS are configured to read the contents and adjust the states as necessary. If the same bits are repeated in successive sensing cycles, CADS’s Analyzer broadcasts distinct $C_{Messages}$ to adapt to a new state. In EDASS, the sensor’s status will change due to the same policy of repeating bits or frequencies in successive packets; however, EDASS also detects the updated event. To identify the location, direction, and intensity of an event, active state sensors send an activating message to all their neighbors. According to the four-state transition methodology, all three of these methods operate on the assumption that the sensor’s state changes from its present state to a new one when the same data, pattern, or frequency is repeated.

State transition model in CADS, EASS, and EDASS

Each sensor can be managed in the state by combining these internal modules. Combining these three elements (microprocessor (MPU), sensing (SU), and communication (RU)) results in a total of sixteen potential states. However, CADS, EASS, and EDASS, as well as HCEE, create and implement only four states. These states meet the requirements of each scheme to ensure full connectivity and never miss any events. These four states are defined based on these internal modules, and it is useful because too many states also degrade system performance. The sensor state is the energy level of the sensors, and it is adjusted according to the contents of the packets. Although the names of the four state models used in these three systems change, the energy of each state remains consistent. In Table 9, all possible states have been shown with their internal components and energy levels. In Fig. 3, the internal components of the sensor are shown with energy levels.

Table 9 All possible states in a single Sensor.

Fig. 3

All possible states for any sensor.

Dynamic and adaptive techniques

The mechanisms of these schemes are dynamic because the network responds appropriately to changes in the network based on content verification without changing the structure of the network. The topology remains active at all times, while the sensors’ states change regularly using a four-state model. Various sensor states regulate network traffic according to network requirements and control sensing, processing, and communication processes. Sensor scheduling enables the system to consume minimal energy by reducing clustering overhead and finally improves energy efficiency. The sensors are strictly configured with state-based sensing; however, the network remains active. Even the sleeping sensor can also remain in such a position that at any time it can switch to the active state without any time lapse. The role of the sensor always determines the state of the sensor, and the new state is adapted using these contents. Sensors’ scheduling is adaptive because the network adapts to changes as sensors transit from one state to another. The network structure may be modified by a single control message or an awakening broadcast message, and the network adapts to these changes dynamically. Table 10 compares three methods based on core parameters.

Table 10 Comparison of CADS, EASS, and EDASS with core parameters.

Methodology of hybrid of CADS, EASS, and EDASS (HCEE)

After discussing all the features of these three schemes with their main core idea, their components, and functional properties, we have derived a hybrid structure of these three schemes. The hybrid is based on the best features of these schemes and a generic structure to follow in any IoT scenario to optimize energy and enhance the performance of the system. In the following section, we have mentioned its main features and technical details.

Energy model of HCEE

There are many energy models used to configure and use energy in sensors in IoT, but in this case, two famous models were used. One is a first-order radio model, and the second is a residual energy model with controlled communication and computation at the sensor level. In the case of HCEE, the second approach with changes is recommended to used. For DPM, the duty cycle is a mechanism to keep some components in active mode while others are in sleep mode. This is always used for energy conservation in sensor IoT. Defined as the ratio of the active time vs total, as follows.

$$\begin{aligned} C_T = C_{(Act)}/C_{(Tot)} \end{aligned}$$

(1)

$$\begin{aligned} C_T = C _{(Act)}/(C _{(Act)} + C _{(SlP)} ) \end{aligned}$$

(2)

where $C_T$, is the duty cycle, $C _{(Act)}$ is the time when the sensor is active, and $C _{(Slp)}$ is the hibernation time. The energy in any sensor can be calculated using these terms of energy consumption.

The energy consumed by the sensor in transmitting $En_{(Tn)}$ and receiving $En_{(Rn)}$. These are the factors of energy when the radio module sends or receives data. The processing energy $En_{(Pr)}$ that is used by the processor in a sensor. The sensing unit consumes energy $En_{(Sn)}$ when it detects sensation.

Sensors typically consist of five parts/modules (Processor, sensing, radio, memory, battery), but energy consumption involves only three main parts: these are Processor $P_r$, Sensing module $S_n$, and Radio module $R_n$, as shown in Fig. 4. Based on these modules, different states are defined.

Fig. 4

Three components of Sensor.

The total flow of current inside the sensor, based on the above three components, can be calculated as follows.

$$\begin{aligned} I_{Tot} = I_{Sn} + I_{Pr} + I_{Rn} \end{aligned}$$

(3)

In the energy consumption in its current form, four states have been defined. Equation (3), is rewritten as follows.

$$\begin{aligned} I_{Tot} (t) = I_{Sn} (t) + I_{Pr} (t) + I_{Rn} (t) \end{aligned}$$

(4)

Current is the flow of charge in unit time; the Eq. (4), is rewritten as follows.

$$\begin{aligned} dQ_{Tot}/dt = d(Q_{Sn})/dt + d(Q_{Pr})/dt + d(Q_{Rn})/dt \end{aligned}$$

(5)

For the charge change over time, the Eq. (5) has been replaced.

$$\begin{aligned} Q = \int _{T1}^{T2} I (t) dt \end{aligned}$$

(6)

Further, it can be replaced and verified as follows.

$$\begin{aligned} Q = \int _{T1}^{T2} I _{Sn} (t) dt + \int _{T1}^{T2} I _{Pr} (t) dt + \int _{T1}^{T2} I _{Rn} (t) dt \end{aligned}$$

(7)

Electric charges in the form of kinetic or potential energy are converted to joules. where the electric circuit’s power is the rate at which energy is maintained as follows.

$$\begin{aligned} P=VI \end{aligned}$$

(8)

For V is the voltage, and resistance R, it can be simplified as follows.

$$\begin{aligned} P=VI= I^2 R = V^2/R \end{aligned}$$

(9)

$$\begin{aligned} E = VI (t) P (t) = V (t) I (t) \end{aligned}$$

(10)

$$\begin{aligned} E = \int _{i}^{t} V (t) I (t) dt \end{aligned}$$

(11)

Equation (11) is the total energy in a sensor used, while Eq. (7) is the individual modules of energy consumed.

Four state model

On the energy consumption in each sub-module of sensors, different states are defined. A total of sixteen possible cases, only four useful states are used in all three schemes. Because energy efficiency influences how quickly states can change if more states are used, these four derived states are advantageous. The four-state approach guarantees complete node connectivity and never overlooks an event. The four-state model is shown in Fig. 5. These states ensure to obtain the dynamic structure of the sensors for energy optimization.

Fig. 5

Four state model.

For sensors to be in an active state (state-1), all three modules are in active mode. The active state can be represented as.

$$\begin{aligned} E_{Tot(i)} = E_{Pr} + E_{Sn} + E_{Rn} \end{aligned}$$

(12)

$$\begin{aligned} \begin{aligned} E_{Tot(Act} = \sum _{j=1}^{n} P_{Pr-state} (j) C_{Pr-state} (j) +\\ \\ + \int _{T1}^{T2} V_{Tn} (m) (I_{Sn-start} (M)) + (I_{Sn-end} (M)) dt\\ + \int _{T1}^{T2} V_{Tn} (M) (I_{Rn-start} (M)) + (I_{Rn-end} (M)) dt \end{aligned} \end{aligned}$$

(13)

For wait state (state-2), two modules: Pr and Sn are active while Rn is in sleep. The energy calculation in the wait state is as follows.

$$\begin{aligned} \begin{aligned} E_{ToT(Wait)} = \sum _{j=1}^{n} P_{Pr-state} (j) T_{Pr-state} (j) + \\ + \int _{T1}^{T2} V_{Tn} (M) (I_{Sn-start} (M)) + (I_{Sn-end} (M)) dt \end{aligned} \end{aligned}$$

(14)

For the sleep state (state-4), all modules are in sleep mode with very little usage of energy. It can be expressed as follows in an equation.

$$\begin{aligned} \begin{aligned} E_{SN(Slp)} = \sum _{k=1}^{m} No_{Pr-change} (k) En_{Pr-change} (k) \end{aligned} \end{aligned}$$

(15)

While for route state (state-3), the path-establishing sensors are used. In this state, the Sn and Rn are inactive, while the Pr is in a sleep mode.

$$\begin{aligned} \begin{aligned} E_{SN(Route)} = \int _{t1}^{t2} V_{Tr} (m) (I_{SU-start} (m)) + (I_{SU-end} (m)) dt\\ + \int _{t1}^{t2} V_{Tr} (m) (I_{RU-start} (m)) + (I_{RU-end} (m)) dt \end{aligned} \end{aligned}$$

(16)

New event detection in HCEE

When a new event is detected or an event is updated, the system reshapes the network structure. All the receiving sensors broadcast a message to all the neighboring nodes and the BS about the new event.

Sensors are deployed in the target area, and after several sensing cycles, they adapt to different states (1–4). States are energy levels used for enhancing the energy efficiency of the IoT, and these are assigned to the roles of sensors in the network. The updated/new event can only be detected by active-state sensors; they just broadcast a message to all other state sensors to become active. When this control message is received, all the sensors are now in an active state and start sensing in a normal way.

Normal operations

After installing the sensors in normal mode, they get updates and follow a schedule in four states. These states are the energy levels, and maximum energy is consumed by active states, while others use moderate and even very little amounts. The main diagram of HCEE is shown in Fig. 6, with different states, CHs, and the BS. There is are scheduling for these sensors, and some of them are sending data, and some of them just hold it. All data is collected by the primary CHs and then forwarded to the master CH. The master CH is then forwarded to BS. Where BS applies different operations on sensor data and analyzes the contents.

Fig. 6

HCEE system working procedure.

Re-scheduling at event detection

For event detection and checking the contents, we have used the four-bit contents and checked the state-based activities. Each of the sensors follows the duty cycling pattern and adapts to new states. Consider a four-bit data scheduling in Table 11 with proper states. Sensors follow this configuration for data detection, find similarities in data, and then change the state as recommended by the state model. In updated event detection, if the contents changed, such as 1111. The active sensor detects it and broadcasts a packet to its neighbors to activate, and also sends a copy to the BS. Other state sensors hear this message on their low-level magnetic sensors, as this is the lowest energy level in the sleeping mode of the sensor module. They hear and become active and start sensing in an active state. All the possible values with four states are mentioned in the table, and sensors have been scheduled accordingly.

Table 11 4-bit data pattern for changing state in HCEE.

HCEE workflow

In each cycle, data are analyzed for traces with previous data, and the same contents are found for consecutive times, changing the state of the sensor. Initially, all sensors are in an active and after three cycles, if the same data is detected, the state is switched to a wait state. In the same way, if the contents are still observed with the same contents, the state is changed to the sleep state. The route state is only used for path established sensors and is a rear state. All this process is systematically arranged in an Algorithm ??.

Algorithm 1

Condensed 4-Bit state transition flow

Evaluation and comparative analysis of CADS, EASS, EDASS, and HCEE

The behavior of these schemes has been experimented with in various settings to check the various parameters. These parameters are energy, message overhead, communication delay, sensor-level processing, and memory requirements. There is a final discussion of the overall performance of all schemes. The main parameters and the concerned metric values are mentioned in Table 12.

Table 12 Basic parameters used in the experimental setup for CADS, EASS, EDASS, and HCEE.

Energy efficiency

The main aim of these three schemes (CADS, EASS, and EDASS) was to schedule and adjust the sensors to minimize energy consumption. Each of the schemes has its own implementation with different requirements. An analyzer is used in CADS for analysis and broadcasting differently $C_{Messages}$ in the backward direction after checking the contents of detected data packets. It is a centralized technique, and all activities are controlled by the analyzer module. It uses the first-order radio model with redundant bits^66,67. Now, the analyzer module sends $C_{Messages}$ to some sensors to change the current state. Which minimizes the flow of bits towards MPU is minimized, and the frequent communication over RU is minimized. EASS, on the other hand, compares the values in the sensor and checks the contents abruptly after detecting on RU. Instead of centralized management in CADS, a distributed approach has been applied for data collection. When implementing EASS for minimal energy consumption, self-organization and node buffer are essential components required. With the detection of an event, it broadcasts an awakening message to all nearby sensors to become active. Upon receiving this message, the sensors in other states become active. This technique saves additional processing MPU and decreases the sense at SU. Finlay, HCEE is analyzed in the same scenario as obtaining these energy values for the same number of rounds with an equal number of sensors. It works like CADS with less energy consumption than EASS and EDASS due to its hybrid and lightweight work operation. For the hybrid scheme, the energy model is a residual energy model with controlled communication, and computation is implemented, and its behavior is mapped on the same graph as shown in Fig. 7. The behavior of each scheme on each round is mentioned in Table 13, which shows that each scheme’s performance differed.

Fig. 7

Energy efficiency of HCEE with CADS, EASS, and EDASS.

Table 13 Comparison of HCEE with CADS, EASS, and EDASS in Terms of energy efficiency.

Based on these values, we have calculated some of the statistical values and conducted some necessary tests on these data. In these experiments, it is proven that HCEE’s mean energy efficiency is significantly different from other baseline mechanisms. HCEE exhibits statistically significant improvement in energy efficiency compared with EASS and CADS, is about p < 0.05. EDASS achieves the highest mean and is statistically stronger with the ANOVA tests. This test confirms that overall group differences are significant, as shown in p = 0.019 and mentioned in Table 14.

Table 14 Statistical significance analysis of energy Efficiency in CADS, EDASS, EASS, and HCEE.

Message overhead

Message overhead is calculated for a measure of all the $C_{Messages}$ and other messages that are used in communication. In CADS, the analyzer module is activated at the BS and is responsible for sending different $C_{Messages}$. Sensors switch states in check of the content and intensity of the detected data packets. Self-configuration is a way to change states by maintaining a check on data content and traffic patterns. The new values are compared with the stored values, and if similarities are found, the state will change accordingly. In EDASS, when a sensor in active-state detects a new or updated event, it transmits a $C_{Message}$ for all neighboring sensors. Creating and broadcasting this $C_{Message}$ creates an extra overhead on the MAC, which leads to extra overhead. Both the CADS and EDASS need extra MAC slots for transmitting $C_{Messages}$. In a hybrid of all these, the message overhead is there because the messages are broadcast when a new event is detected, although there are very few messages from the base station to the leaf sensors. However, there is still an overhead of control messages in EDASS and HCEE. In CADS, the analyzer is responsible for these messages, while in EDASS, the active sensor transmits these messages. Compared with performance here, 100 sensors in many rounds allowed mapping the behavior of these schemes in Fig. 8, and their statistical values are mentioned in Table 15.

Fig. 8

Message overhead at CH in HCEE with CADS, EASS, and EDASS.

Table 15 Statistical summary for CADS, EDASS, EASS, and HCEE.

Statistical evaluation shows that HCEE exhibits the lowest message overhead (18,225.68), outperforming CADS (17,097.00), EDASS (15,490.00), and EASS (13,890.00). The overall difference between each scheme was statistically significant (F(3, n – 4) = 4.9, p = 0.011). All these values are shown in Table 16.

Table 16 Pairwise t-Test and ANOVA results for CADS, EDASS, EASS, and HCEE.

The message overhead at the CH is higher in CADS than in EDASS and EASS. Due to the awakening messages, the EDASS and HCEE also create some extra overhead. The message overhead at CH is lower in EASS than in the other two schemes because there is no $C_{Messages}$. All three schemes with hybrid (CADS, EASS, and EDASS) have been tested in 100 sensors with 1000 rounds; calculate the number of messages at CH. After 200 rounds, CADS starts to increase a little and becomes uniform around 17,500, and abruptly declines after 700. As shown in Fig. 9, the decrease is very small and continuous, which means that the experiments ended if extended to 300 rounds. Table 17, with statistical values of increasing efficiency in message overhead of about 7.37% in HCEE compared to all three schemes.

Fig. 9

Messages at base station.

Table 17 Statistical summary of messages at base station at HCEE, CADS, EASS, and EDASS.

Communication delay

Delay in communication $(D_{Communication})$ is calculated as: $D_{Communication}$ = ($N_{Packets}$ + $S_{hop}$) * ($L_{Packets}$ / $R_{Rate}$), in which “N” is the number of packets transmitted, $S_{hop}$ is number of hop counts, $L_{Packets}$ is the size of packet and $R_{Rate}$ is the rate at which the packets were transmitted. We computed the delay after experimenting with CADS, EASS, and EDASS with HCEE, for 100 sensors over 1000 rounds. Figure 10 shows a graph that plots the behavior of each scheme. Due to the lower message overhead, EASS is a lighter scheme than HCEE, EADSS, and CADS. Only the node uses the self-organization mechanism to reply after reviewing the information. After 300 rounds, the average energy consumption of the network begins to stabilize or decrease when the node states are uniform. The self-configuration approach reduces the EASS latency. There is no additional overhead of $C_{Messages}$ compared to CADS and EDASS. In CADS, the delay increases up to 640 rounds before beginning to level out and then decreases. To keep active and sleep sensors in a dynamic structure, the analyzer transmits various $C_{Messages}$ to all sensors in the network. CADS follows backward communication, and the analyzer transmits different messages in the backward direction. The extra overhead is created by these $C_{Messages}$ from the analyzer, and all of these create a delay in communication. The communication latency in the case of EDASS is greater than that of EASS but lower than that of CADS and HCEE. As seen in Fig. 10, the network as a whole is also impacted by the total latency of these communications. The start decreases, and the delay increases to 550 rounds in this graph. The statistical values of CADS, EASS, EDASS, and HCEE for different terms are calculated in Table 18.

Fig. 10

Communication delay between HCEE and CADS, EASS, EDASS.

Table 18 Communication delay between HCEE and CADS, EASS, EDASS.

Comparing the statistical values, CADS achieves the lowest mean communication delay (0.5782 s) with better latency performance. As shown in Table 19, the ANOVA (p = 0.009) shows that differences between CADS, EASS, EDASS, and HCEE are statistically significant.

Table 19 Inferential statistical analysis of communication delay (pairwise t-tests and ANOVA).

Processing at the node level

Energy is consumed by the processor to perform several functions, such as channel filtering, frame synchronization, modulation, demodulation, bit conversion across various processing units, and stream control. Some other tasks are buffering, CRC, and applying security rules, with some aggregation and comparison tasks. We tested all three schemes with a hybrid scheme with an Intel Core i5-14600KF processor (20 cores) 3.5 GHz, with 32 GB RAM, and supported by a dedicated NVIDIA RTX 5070 Ti GPU. Considering the operational processes of all three of these schemes (CADS, EASS, and EDASS) with the hybrid scheme, the majority of operations (comparison) are carried out at the node level, with EASS and EDASS being more focused at the node level. In CADS, all additional processes on a sensor, such as remembering signatures, comparing strings, and computing signatures, are omitted. To remember, compare, and compute signatures in detected packets, EASS and HCEE required some additional node-level computation. This creates additional overhead, which consumes a good amount of energy. All three schemes are experimented with in the same terrain, and their performances are mapped in Fig. 11. EASS utilizes more MPU than CADS because it concentrates more on node-level processes. In EASS, all activities are carried out at the node level (MPU), whereas CADS is simply sensed and forwarded to the base station for additional analysis. These experiments demonstrate that EASS uses a higher percentage of MPU cycles.

Fig. 11

Processing at node level in CADS, EASS and EDASS.

Memory requirements

Both EASS and EDASS use compression at the sensor level. For this, they need an extra buffer at the sensor level. However, EASS employs self-organization techniques and, from the content analysis, the sensor adjusts to new states and changes states using the same values. In contrast, the analyzer module in CADS broadcasts various $C_{ Messages}$ to sensors, compares values, and records detected data. As a result, more memory is required for self-configuration in EASS but not in CADS. For implementation, we have used the commercially available CC2420 product with a defined buffer size and other characteristics in MATLAB. According to these requirements, the transmitting side’s buffer size is 128 bytes FIFO, and the receiving side’s is 128 bytes FIFO. Different packet sizes range from eight bytes to 256 bytes (frame size). It can occasionally fall between 24, 28, 30, 40, 60, 64, 90, 120, 128, 256, and 512 bytes. In the CC2420 specs, take into account various packet sizes for 128-byte memory in the receiving FIFO. EASS and EDASS are implemented in 100 sensor nodes with four state models and 0.5 joules of energy each for performance analysis. Figure 12 plots the behavior of the terrain after it has been tested for various packet sizes with varying events. The results show that a packet buffer of 128 bytes takes 43 s longer than a buffer of smaller sizes. In single event detection, a 64-byte buffer fills in 30 s, a 32-byte buffer fills in 19 s, and a 16-byte buffer fills in 13 s. In contrast, 10 distinct events fill the buffer at various times in the same situation, requiring 128 bytes of buffer space for 30 s, 64 bytes for 19 s, 32 bytes for 12 s, and 16 bytes for 8 s. The lines stabilize after the 30th event, and all buffers are filled at different times with little variation. All of these statistics indicate that CADS needs a smaller memory buffer than EASS, which uses more memory. The Fig. 13 shows all experimental values for filling buffers in bytes in CADS, EASS, and EDASS. The comparison of the HCEE with all these is shown in Table 20, where it shows near CADS and requires more buffer than EASS and EDASS.

Fig. 12

Different buffer sizes and required time.

Fig. 13

Filling buffers in bytes in HECC, CADS, EASS, and EDASS.

Table 20 Average buffer size (in bytes) across different numbers of nodes for CADS, EASS, EDASS, and HECC schemes.

Significance of the results

The evaluation of HCEE with CASD, EASS, and EDASS for various metrics provides a comprehensive understanding and comparative analysis.

Energy efficiency: Analysis of HCEE’s energy efficiency performance showed that it performed better than CADS (17,097.00), EDASS (15,490.00), and EASS (13,890.00), with the highest mean energy efficiency (18,225.68). Based on these values, the ANOVA test exhibits statistically significant differences in HCEE with CASD, EASS, and EDASS. While Pairwise t-tests, HCEE performed better than CADS ($p=0.047$) and EASS ($p<0.001$), although the difference with EDASS was negligible ($p=0.062$). All these values suggest that HCEE provides better energy utilization and consumes energy when needed by applying scheduling at the sensor level.

Message overhead: HCEE performs better in terms of message overhead, and it achieves good performance with the lower overhead (18,225.68), compared with CADS (17,097.00), EDASS (15,490.00), and EASS (13,890.00). This behavior of HCEE reveals the significant improvement over CADS ($p=0.047$) and a highly significant advantage over EASS ($p<0.001$). These statistics confirm the effectiveness of HCEE in minimizing communication redundancy.

Communication delay: In communication delay, HCEE is quite poor compared to CADS. This is because CADS takes less time in broadcasting the messages and reaches to the required place BS. CADS has achieved the lowest mean delay (0.5782 s) compared to HCEE (0.9256 s), EASS (1.2729 s), and EDASS (1.5224 s). These values are further used in statistical analysis in which the differences were statistically significant, shown as $F(3, n-4)=5.1$, $p=0.009$. In pairwise comparisons, CADS performed better than EASS and EDASS with significance ($p<0.01$). On the other hand, the difference between CADS and HCEE is $p=0.042$, which is not high. All this means that CADS exhibits faster communication cycles.

Buffer utilization: In buffer utilization at sensors, CADS needs a higher buffer to keep all values in sensor memory with 84.00, while HECC needs 74.70. The other two, EDASS needs 69.33 and EASS needs 60.17 to fulfill in the required time and keep the slices of data for comparison. Using the pairwise t-tests, all schemes showed no statistically significant differences ($p>0.1$).

From all these statistics, HCEE maintains a stable buffer management policy in maintaining a well-balanced trade-off between energy efficiency, message overhead, and communication delay. The ANOVA results and observed p-values provide compelling evidence that HCEE performs better than CADS, EASS, and EDASS.

Discussion

The main goal was to enhance the energy of IoT sensors in CADS, EASS, HCEE, and EDASS by using scheduling, which regulates the internal operations of a sensor. With implementation, a sensor can deliver important packets with fewer sensing cycles and avoid producing redundant or duplicate packets. This lowers the overhead related to re-clustering by generating less traffic in control messages and data packets. For each sensor, a constant energy is always required, and we always need an energy-conscious method to enhance the energy efficiency. In most literature, this issue is solved by using upper-layer policies and never checking the lower sensor-level details. If the sensing capabilities of sensors are controlled and adjusted according to the role of the network, it can greatly affect the system’s efficacy. For this, a data link layer technique is needed instead of upper-layer implementations. These are the main aims of this work, and it has followed the scheduling at the sensor level. Applying active/sleep techniques with adaptive and dynamic node scheduling can improve energy efficiency. The main basis was a content-based, dynamic, adaptable, and real-time data inspection to define roles and sensor states. Adaptive management is applied by applying these states, and the sensors are adjusted accordingly.

Figure 14 illustrates the overall impact of the three schemes on the overall dependence of one factor on another. Sensor states are used to regulate extensive sensing in the initial phase. Fewer bits are fetched to the microprocessor for processing when sensing is reduced. When processing is regulated, communication is optimized, and only the necessary packets are sent across the network. CADS is a centralized technique where control messages are sent, and the contents of sensed data packets are examined at the BS/Analyzer. Control messages play a crucial role in altering sensor states. Sensors are adjusted using these messages, which improves system energy efficiency and extends network lifetime. In EASS, a sensor can adapt to any other state based on its role using the Four-State Model. However, the contents are analyzed at the sensor level using a self-configuration method. It applies the state-based procedures at the sensor for energy efficiency. The EDASS is basically designed for detecting new or updated events. The state of the sensor changes as a result of the detection of a new event. The active-state sensor sends a message to all other sensors after detecting an updated event. While HCEE combined all the best features of these three schemes and it is a hybrid. It combines unique advantages for energy efficiency, content analysis, event detection, and states. These features are combined in one scheme to increase overall system performance in adaptive scheduling.

Fig. 14

Overall impact of CADS, EASS, and EDASS.

Conclusion and future directions

IoT sensors constantly require energy-efficient protocols and energy-conscious policies to minimize energy consumption to enhance system performance. To address the issue related to energy enhancement in IoT sensors, this paper explains various techniques that have been implemented at the sensor level and operate at the data connection layer. By using active/sleep techniques with adaptive and dynamic node scheduling, these schemes improve energy efficiency. The ultimate objective was to provide content-based, dynamic, adaptable, and real-time solutions. In this paper, three methods were used to reach these goals: (1) to analyze the traffic patterns and contents by an analyzer module, (2) to check the contents in the sensor buffer and compare them with the signature of the data, and (3) to change state with new event detection. The first scheme (CADS) is a centralized method in which various control messages are broadcast, and contents are examined at the base station (Analyzer). These messages are very crucial in altering states. Sensors are switched between states using various control messages, which improves system energy efficiency and extends network lifetime. In the second approach (EASS), a sensor can access any unique state that uses the Four-State Model. However, contents are verified at the sensor node by the self-configuration method. Enhances energy efficiency by applying state-based procedures at the sensor. In the third scheme (EDASS), the system works on new or updated event detection. The state of the sensor changes as a result of the detection of a new event. Using this scheme, we can detect the direction and intensity of an event. The active-state sensor transmits a message to all other sensors after detecting an updated event. In the HCEE, it is basically a hybrid of all these schemes with combines the most effective features of CADS, EASS, and EDASS. These models present some unique advantages for energy efficiency, content analysis, event detection, and states. All of these features are combined in one scheme to increase overall system performance in adaptive scheduling. Similar to EDASS, HCEE uses a node-level content analysis method for content analysis, which ensures that only significant data are processed and sent. Like CADS, which manages role assignments and decision logic through a centralized policy, HCEE implements it for enforcement and content checking. For event detection, it follows the policy of EDASS, which works for the quick detection of events.

In the future, our goal is to integrate the best features of all three schemes: CADS, EASS, and EDASS. These models may be expanded to the real packet size and data length, even if they are only tested with sample data up to four bits for ease of use. In addition, the frequency distribution and recurring patterns should be examined using the existing model. In an energy grid where all sensors work together to gather data and control operations based on traffic and circumstances, EASS can be helpful. In the energy grid, additional traffic and functions can be evenly allocated across sensors. These approaches will be explicitly tested using validation and verification methods such as Petri nets. A statistical and quantitative analysis will be conducted to assess their validity and the behavior of these schemes; they will be tested in a real-time test-bed scenario.