Network relay scheduling

Abstract

Techniques are disclosed for accurately predicting the occurrence of abnormal sensor readings within a sensor network and advantageously using these predictions to limit the amount of power used by relay nodes within the sensor network. Some examples analyze the spatial and temporal characteristics of abnormal sensor readings to predict future occurrences. In these examples, relay nodes operate in a reduced-power mode for time periods when abnormal sensor readings are not predicted to occur. Also, in these examples, only relay nodes in the path between the sensor where the abnormal reading was predicted and a gateway of the sensor network operate in full-power mode. This feature allows other relay nodes to remain in a reduced-power mode even when abnormal sensor readings are predicted elsewhere in the sensor network.

Description

Background Art

[0001] Spatially distributed wireless sensor networks (WSNs) are widely deployed in Internet of Things (IoT) applications, such as environmental monitoring systems, traffic and parking monitoring systems, and utility monitoring systems. One functional requirement of these systems is to capture abnormal or anomalous readings in the WSN and report them to a gateway within a time-limited wait time period. This wait time period is defined by the specific IoT application. For example, in some cold chain logistics applications, sensor nodes must report abnormal temperature readings to the gateway within 1 minute. Short wait time periods such as these significantly increase power consumption within the WSN because relay nodes within the WSN along the routing path to the gateway must operate continuously to receive, send, or route abnormal temperature readings to meet the wait time period. Continuously operating relay nodes can consume significantly higher power, for example, in some configurations, consuming nearly four times the power of edge nodes in a mesh WSN. For this reason, relay nodes can dominate the WSN power draw and shorten its operating life. BRIEF DESCRIPTION OF THE DRAWINGS

[0002] Figure 1 is a block diagram illustrating a sensor network configured according to an embodiment of the present disclosure.

[0003] Figure 2 is a flowchart illustrating an abnormality prediction process according to an embodiment of the present disclosure.

[0004] Figure 3 is a flow chart illustrating a duty cycle scheduling process according to an embodiment of the present disclosure.

[0005] Figure 4 is a flowchart illustrating a model training process according to an embodiment of the present disclosure.

[0006] Figure 5 is a block diagram of a computing device that may be used to implement various components of a sensor network according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0007] The systems and methods disclosed herein accurately predict the occurrence of abnormal sensor readings within a sensor network and advantageously use these predictions to limit the amount of power used by relay nodes within the sensor network. In some embodiments, the systems and methods disclosed herein analyze the spatial and temporal characteristics of abnormal sensor readings to predict future occurrences. In these embodiments, the relay nodes operate in a reduced power mode for time periods during which abnormal sensor readings are not predicted to occur. Likewise, in these embodiments, only relay nodes in the path between the sensor for which the abnormal reading is predicted and the gateway of the sensor network operate in full power mode. This feature allows other relay nodes to remain in reduced power mode even when abnormal sensor readings are predicted elsewhere in the sensor network. These advantages enable the entire sensor network to consume less power than a conventional sensor network in which relay nodes operate continuously in full power mode.

[0008] Some embodiments relate to a sensor network comprising sensor nodes, relay nodes, and a gateway. The sensor nodes include environmental sensors that acquire sensor readings to measure one or more characteristics of an environment of the sensor network. The sensor nodes transmit data describing the acquired sensor readings to the relay nodes. The relay nodes receive the data and transmit the data to the gateway. The gateway connects the sensor network to one or more other communication networks and transmits the data and / or a summary of the data to the other communication networks.

[0009] In some embodiments, each of the sensor nodes includes a modeling component and each of the relay nodes includes a scheduling component. These modeling components and scheduling components work in combination to ensure that all abnormal sensor readings are predicted, acquired, and relayed to the gateway. More specifically, the modeling component is preconfigured to perform a modeling process that predicts future abnormal sensor readings. In the event that the modeling process predicts future abnormal sensor readings, the modeling process generates one or more duty cycle requests and transmits the one or more duty cycle requests to one or more relay nodes in the path between the sensor node and the gateway. In some embodiments, these duty cycle requests include information describing the duty cycle and duration, which is requested by the sensor node to enable the one or more relay nodes to be in full power mode at the time the abnormal sensor readings are acquired.

[0010] The modeling process can be based on any number of prediction techniques, such as time series methods, artificial intelligence methods, and simulation methods. For example, in some embodiments, the modeling process can implement polynomial curve fitting, neural networks, and / or decision trees. In at least one embodiment, the modeling process implements a regression method. In some embodiments, the modeling component periodically reconfigures the modeling process to improve the modeling process by comparing actual sensor readings with previously predicted sensor readings and adjusting the modeling process to reduce the error between the actual sensor readings and the sensor readings predicted by the modeling process.

[0011] In some embodiments, the scheduling component is configured to receive and process duty cycle requests to implement the duty cycle and schedule requested therein. For example, when executed according to this configuration in at least one embodiment, the scheduling component receives the duty cycle request, parses the request to identify the requested duty cycle and duration, and changes the configuration information of the relay node that executes the scheduling component to implement the requested duty cycle and duration. In doing so, the scheduling component may change the operation of the relay node from a reduced power mode to a full power mode for the requested duration, thereby enabling the relay node to receive, process, and forward any predicted abnormal sensor readings, if they occur. At the expiration of the requested schedule, the scheduler restores the configuration information of the relay node to its previous state. This reconfiguration causes the relay node to operate in its reduced power mode, thereby saving power relative to a conventional, always-on relay node.

[0012] Other aspects, embodiments, and advantages of these example aspects and embodiments are discussed in detail below. In addition, it should be understood that the foregoing information and the detailed description that follows are merely illustrative examples of the various aspects and embodiments and are intended to provide an overview or framework for understanding the nature and characteristics of the claimed aspects and embodiments. References to "an embodiment," "other embodiments," "an example," "some embodiments," "some examples," "alternative embodiments," "various embodiments," "one embodiment," "at least one embodiment," "another embodiment," "this and other embodiments," etc. are not necessarily mutually exclusive and are intended to indicate that a particular feature, structure, or characteristic described in connection with that embodiment or example may be included in at least one embodiment or example. . The appearances of such terms herein do not necessarily all refer to the same embodiment or example. Any embodiment or example disclosed herein may be combined with any other embodiment or example.

[0013] Moreover, the wording and terminology used herein are for descriptive purposes and should not be considered restrictive. Each example, embodiment, component, element or system and method action cited in the singular herein may also include plural embodiments, and any embodiment, component, element or action cited in the plural herein may also include only singular embodiments. Reference in the singular or plural form is not intended to limit the currently disclosed system or method, and its components, actions or elements. "Include," "contain," "have," "include," "involve," and their variations used herein are intended to cover the items listed thereafter and their equivalents and additional items. Reference to "or" may be interpreted as inclusive, so that any term described using "or" may indicate one, more than one, and all of the terms described. Additionally, in the event of inconsistent term usage between this document and the documents combined by reference, the usage in the combined reference should be considered to supplement the usage of this document; for irreconcilable inconsistencies, the term usage in this document shall prevail.

[0014] General Overview

[0015] The power consumption of relay nodes operating in full power mode is often wasteful because WSNs typically operate under normal, non-abnormal conditions. Although previous work has explored the use of custom hardware features (such as wake-up radios) to reduce relay node power consumption, this previous work has failed to exploit the properties of sensor data fields to save WSN power.

[0016] Therefore, and in accordance with at least some embodiments disclosed herein, a sensor network implements a predictive process based on the characteristics of sensor data fields generated by the sensor network. In these embodiments, the sensor network intelligently controls the duty cycle of relay nodes to reduce the sensor network's power consumption while meeting stringent low-latency anomaly reporting requirements. These predictive processes may include learning-based models that exploit the data characteristics of the sensor fields and, therefore, predict anomalies in the sensor network before any anomalous sensor readings occur. In these embodiments, the duty cycle of relay nodes in the sensor network is tuned to the data stream predicted to be acquired by the sensor nodes. As part of this tuning, the relay nodes are dynamically set to a longer reduced-power mode schedule, which significantly reduces the power draw of the relay nodes' idle listening. As another part of this tuning, the relay nodes are dynamically set to a shorter full-power mode schedule to cover predicted anomalous sensor readings, preventing anomalous sensor readings from going unreported. Additionally, at least some embodiments disclosed herein are software-based and therefore do not require any specialized hardware implementation within the sensor network.

[0017] System Architecture

[0018] Figure 1 1. A sensor network 100 is shown including a gateway 102, a plurality of relay nodes 104A to 104N, and a plurality of sensor nodes 106A to 106N. The nodes of the sensor network 100 may include, for example, one or more IoT devices. Figure 1 As shown in FIG, each of the relay nodes 104A to 104N includes a memory 108, at least one interface 110, at least one processor 112, and at least one scheduler 114. Each of the sensor nodes 106A to 106N includes a memory 108, at least one interface 110, at least one processor 112, and at least one modeler 118.

[0019] like Figure 1 As shown in , each memory 108 may include volatile and / or non-volatile data storage (e.g., read-only memory, random access memory, flash memory, magnetic / optical disks, and / or some other computer-readable and writable media) that is readable and / or writable by one of the processors 112. Memory 108 is sized and configured to store a program that can be executed by one of the processors 112 and at least some of the data used by the program during execution. Each of the processors 112 includes various computational circuits, such as an arithmetic logic unit and register memory, that can execute instructions defined by the instruction set supported by the processor 112. Each of the processors 112 may include a single-core processor, a multi-core processor, a microcontroller, or some other data processing device. Each of the interfaces 110 includes communication circuitry, such as a wired or wireless Ethernet port, that enables one of the processors 112 to communicate with the other processors 112 in the sensor network 100. Each of the sensors 116 includes analog and / or digital circuitry that can sample the operating environment near the sensor 116 and measure some characteristics of the operating environment. For example, each of the sensors may include a temperature sensor, a barometer, an accelerometer, or some other sensor.

[0020] In some embodiments, each of the relay nodes 104A to 104N is configured to operate in at least a full power mode and a reduced power mode. When operating in the full power mode, the relay node monitors, receives, and processes inbound communications. When operating in the reduced power mode, the relay node does not monitor, receive, and / or process at least some inbound communications. Conversely, when operating in the reduced power mode, the relay node performs the necessary processes for it to identify when to warrant switching to the full power mode. These processes may include monitoring the interface 110 for special types of inbound communications and / or simply executing a timer that expires when the relay node should switch to the full power mode.

[0021] like Figure 1As shown in , each of the schedulers 114 includes software and / or hardware that can be executed by one of the processors 112 and is configured to receive and process duty cycle requests from one or more of the sensor nodes 106A to 106N. In some embodiments, these duty cycle requests include information describing the duty cycle for the relay node requested by the sensor node. For example, in some embodiments, the duty cycle request includes one or more fields defining the requested duty cycle. Examples of duty cycles that can be requested include continuous duty cycles and periodic duty cycles. Relay nodes that perform according to a continuous duty cycle operate in full power mode or reduced power mode. Relay nodes that perform according to a periodic duty cycle alternate between full power mode and reduced power mode according to a specified time period (e.g., 1 minute in full power mode followed by 4 minutes in reduced power mode). Other types of duty cycles (e.g., non-periodic duty cycles) are contemplated and the embodiments described herein are not limited to specific duty cycles.

[0022] In some embodiments, the duty cycle request also includes at least one field defining the total duration of the requested duty cycle. This duration field stores information specifying the total time period that the requested duty cycle should be executed before returning to the previous or default duty cycle. In some embodiments, the default duration included in the duty cycle request is equal to the time period within which the sensor node has predicted future sensor readings. This time period is referred to herein as the "prediction horizon". Additionally, in at least one embodiment, the duty cycle request does not include a field storing actual duty cycle information as described above, but instead includes an identifier of a predefined duty cycle stored in the memory 108 of the relay node. This predefined duty cycle may include information similar to that of the duty cycle request described above.

[0023] In some embodiments, in receiving and processing duty cycle requests, each of the schedulers 114 parses each request, identifies the requested duty cycle, and manipulates configuration data stored in the memory 108 of the relay node where the scheduler 114 resides to set the requested duty cycle. In some embodiments where the duty cycle request includes a duration field, each of the schedulers 114 is configured to identify the duration field and manipulate the configuration data for the requested duration to implement the requested duty cycle. An example of a scheduling process performed by each of the schedulers 114 in some examples is described below with reference to Figure 3 is further described.

[0024] like Figure 1As shown in , each of the modelers 118 includes software and / or hardware that can be executed by one of the processors 112 and is configured to predict abnormal sensor readings within a prediction time domain (e.g., 4 minutes from the current time). In some embodiments, in predicting abnormal sensor readings, each of the modelers 118 performs a modeling process. This modeling process can be trained using actual or synthetically derived data as a reference for the following Figure 2 and Figure 5 The training data may include data representing various abnormal scenarios (e.g., host spot abnormalities, door opening abnormalities, air conditioner malfunction abnormalities, etc.). Although various modeling processes may be used, in at least one embodiment, each of the modelers 118 implements a regression model to predict future temperature readings from the sensor 116. Equation 1 shows an example regression model used in some embodiments.

[0025] T t+K =c0+c1*T t +c2*T t-1 +c3*T t-2 +...+c P+1 *T t-p Equation 1

[0026] In Equation 1, T t 、T t-1 、T t-2 …T t-P is the historical temperature reading and the coefficients C0, C1, C2, C3, C4..., C P is a parameter calculated during the execution of the training process to minimize the error between the predicted temperature and the true temperature in the training data. Equation 1 predicts the temperature at time t+K, where K represents the prediction horizon. Figure 2 、 Figure 4 and Figure 5 An example of such a training process is further described.

[0027] In some embodiments, each of the modelers 118 is configured to declare an anomalous sensor reading imminent if the modeling process predicts that one or more sensor readings meet one or more predefined criteria. Examples of one or more sensor readings that meet the predefined criteria include a sensor reading exceeding a threshold, a plurality of sensor readings acquired within a predefined time window exceeding a threshold, and a plurality of sensor readings within the predefined time window having a variance exceeding a threshold.

[0028] In some embodiments where the modeling process predicts future temperature readings, each of the modelers 118 is configured to generate a temperature reading at a future temperature predicted by the modeling process (e.g., T t+K ) exceeds a threshold value (e.g., 15 degrees Celsius). In other embodiments, one or more of the modelers 118 are configured to declare an abnormal temperature reading imminent if the future temperature predicted by the modeling process exceeds a threshold value adjusted by a relaxation factor. In these embodiments, the relaxation factor is designed to eliminate false negatives (i.e., not declaring an abnormal temperature reading imminent when it occurs) at the expense of generating false positives (i.e., declaring an abnormal temperature reading imminent when it has not occurred) at the expense of generating false positives (i.e., not declaring an abnormal temperature reading imminent when it has occurred). The value of the relaxation factor can be determined empirically and can be equal to, for example, a boundary set at two standard deviations from the mean sensor reading value.

[0029] In some embodiments, each of the modelers 118 is configured to generate and transmit one or more duty cycle requests (as described above) in response to declaring an impending abnormal sensor reading. These duty cycle requests can be addressed to relay nodes on the path between the sensor node declaring the impending abnormal sensor reading and the gateway 102. In some embodiments, these duty cycle requests include requested duty cycles that place the relay nodes processing the request into their full power mode during a time period that covers the predicted timing of the impending abnormal sensor reading.

[0030] In some embodiments, each of the modelers 118 is configured to perform a training process that refines the modeling process using actual sensor readings as they become available for comparison with sensor readings previously predicted by the modeling process. Figure 3 An example of such a training process is further described. Additionally, further examples of the processes performed by each of the modelers 118 according to the configuration described above are described below in connection with Figure 2 and 4 is provided.

[0031] Although Figure 1 The modeler 118 is shown as residing within the sensor node, but in other embodiments, a single modeler 118 resides within and is executed by the gateway 102. Additionally, it should be understood that the gateway 102, one or more of the relay nodes 104A through 104N, and one or more of the sensor nodes 106A through 106N may utilize computing devices such as those described below with reference to Figure 5 The computing device described is implemented.

[0032] method

[0033] Some embodiments disclosed herein perform an anomaly prediction process, such as in Figure 2 The abnormality prediction process 200 shown in FIG. The abnormality prediction process 200 may be performed by a sensor network (such as the sensor network shown in FIG. Figure 1 The actions performed by the anomaly prediction process 200 collectively build, deploy, and operate a sensor network that conserves power by predicting the occurrence of anomalies and controlling power usage within the sensor network to match these predicted occurrences.

[0034] like Figure 2 As shown in FIG, the anomaly prediction process 200 begins in action 202, where a computing device (such as the following reference Figure 5 The computing device 500 described further uses baseline data to build a process for modeling sensor readings. The baseline data may include historical sensor readings and / or sensor readings generated by simulation synthesis. For example, in an embodiment where the environmental characteristic being monitored by the sensor network includes temperature, the baseline data may include historical sensor readings and / or data generated by a computational fluid dynamics simulation of the environment in which the sensor network will be installed. In some embodiments, the computing device uses regression analysis to build the modeling process. For example, the computing device can build Equation 1 by generating coefficients to minimize the error between the future temperature predicted by Equation 1 and the temperature included in the baseline data. After the modeling process is built, it is installed in each of the sensor nodes of the sensor network as a modeler (e.g., modeler 118).

[0035] In act 204, the sensor network is deployed to its designated environment, and in some embodiments, the modeling process is further trained using actual sensor readings obtained from the environment. For example, in embodiments where the environmental characteristic being monitored by the sensor network includes temperature, the coefficients of Equation 1 may be further adjusted to minimize the error between the future temperature predicted by Equation 1 compared to the temperature included in the actual temperature sensor readings obtained from the environment.

[0036] In act 206, the sensor network is put into operation and at least one modeler (e.g., a modeler housed within sensor node 106A) performs a modeling process to predict abnormal sensor readings. In act 208, the modeler determines whether abnormal sensor readings are imminent. If so, the modeler performs act 210. Otherwise, the modeler performs act 212.

[0037] In act 210, the modeler generates and transmits duty cycle requests to one or more relay nodes (e.g., one or more of relay nodes 104A through 104N) within a path between the sensor node hosting the modeler and a gateway (e.g., gateway 102). These duty cycle requests include duty cycles that cover the predicted timing of the impending anomalous sensor reading.

[0038] In action 212, the modeler determines whether to continue execution. For example, the modeler may check communications from the operating system executing on the sensor node to determine whether a shutdown has been requested. If the modeler determines that execution should continue, the modeler returns to action 206. Otherwise, the modeler terminates the anomaly prediction process 200.

[0039] Some embodiments disclosed herein implement a duty cycle scheduling process, such as Figure 3 The duty cycle scheduling process 300 shown in FIG. The duty cycle scheduling process 300 may be performed by a relay node (such as the one referenced above). Figure 1 The actions performed by the duty cycle scheduling process 300 collectively enable a relay node to receive data from one or more sensor nodes (such as those described above) in addition to the relay node being predicted to receive data from one or more sensor nodes. Figure 1 The relay node executes in a reduced power mode to save power except during time periods when any of the described sensor nodes 106A to 106N receives abnormal sensor readings. During these time periods, the relay node executes in a full power mode as directed by the one or more sensor nodes via one or more duty cycle requests.

[0040] As in Figure 3 As shown in , the duty cycle scheduling process 300 begins in action 302, where the relay node receives a duty cycle request from a sensor node via an interface (e.g., interface 110 of relay node 104A). In action 304, a processor of the relay node (e.g., processor 112) parses the duty cycle request to determine the requested duty cycle and duration. For example, the processor may parse the duty cycle request to identify a periodic cycle during which the relay node operates in full power mode for 5 minutes. In action 306, the relay node executes the requested duty cycle. The action may include the processor changing configuration information stored in a memory (memory 108) of the relay node to implement the requested duty cycle.

[0041] In action 308, the processor determines whether the duration of the currently executed duty cycle has expired. If so, the processor performs action 310. Otherwise, the processor performs action 304. In action 310, the processor restores the relay node to its previous or default duty cycle, and the duty cycle scheduling process 300 ends. In some embodiments, within action 310, the processor changes the configuration information to implement the previous or default duty cycle. For example, within action 310, the processor may restore the duty cycle to a continuous mode operating at a reduced power.

[0042] Some embodiments disclosed herein perform a model training process, such as Figure 4 The model training process 400 shown in FIG. The model training process 400 may be performed by a sensor node (such as the one referenced above). Figure 1 The actions performed by the model training process 400 collectively enable the sensor nodes to refine the modeling process used to predict abnormal sensor readings.

[0043] As in Figure 4 As shown in , the model training process 400 begins in act 402, where the sensor node obtains sensor readings from a sensor (e.g., sensor 116). In act 404, a processor of the sensor node (e.g., processor 112) combines the sensor readings into a set of historical sensor readings stored in a memory of the sensor node (e.g., memory 108). In act 406, the processor refines the modeling process used to predict abnormal sensor readings and the model training process 400 ends. In some examples, the refinement process is performed periodically during operation of the sensor node and calculates parameters of the modeling process (e.g., coefficients in a regression model, links and weights in a neural network model, etc.) that minimize the error between the predicted sensor readings and the actual sensor readings obtained during operation and stored in the set of historical sensor readings.

[0044] Each of processes 200 to 400 depicts a specific sequence of actions in a specific example. The actions included in these processes may be performed by or using one or more computing devices of a special configuration as discussed herein. Some actions are optional and thus may be omitted according to one or more examples. Additionally, the order of the actions may be changed, or other actions may be added, without departing from the scope of the systems and methods discussed herein. In addition, it will be understood that optimization techniques (such as power-efficient data collection and load balancing (e.g., via compressed sampling)) may be used in combination with other processes described herein without departing from the scope of this disclosure.

[0045] computing devices

[0046] Figure 5 A computing device 500 is shown that can be used to implement various components of a sensor network as described herein. As shown, the computing device 500 includes memory 502, at least one processor 504, and at least one interface 506. While the specific types and models of these components may vary between computing devices, it will be understood that each computing device includes a processor, memory, and an interface.

[0047] Interface 506 includes one or more physical interface devices (such as input devices, output devices, and combined input / output devices) and a software stack configured to drive the operation of the device. The interface device can receive input or provide output. More specifically, the output device can provide information for external presentation and the input device can receive information from an external source or generate information. Examples of interface devices include keyboards, mice, trackballs, microphones, touch screens, printing devices, display screens, speakers, network interface cards, environmental sensors, and the like. Interface devices allow programmable devices to exchange information and communicate with external entities such as users and other systems.

[0048] The memory 502 includes volatile and / or non-volatile (i.e., non-transient or non-transient) data storage that can be read and / or written by the processor 504. The memory 502 stores programs and data used or manipulated during the operation of the computing device 500. The program stored in the memory 502 is a series of instructions that can be executed by the at least one processor 504. The memory 502 may include relatively high-performance data storage, such as registers, caches, dynamic random access memory, and static memory. The memory 502 may further include relatively low-performance, non-volatile, computer-readable and / or writable data storage media, such as flash memory or optical or magnetic disks. Various embodiments may organize the memory 502 into special, and in some cases unique, structures to store data that support the components disclosed herein. These data structures may be specifically configured to save storage space or increase data exchange performance and may be resized and organized to store values ​​for specific data and data types.

[0049] In some embodiments, to implement and / or control specialized components, the processor 504 executes a series of instructions (i.e., one or more programs) that result in manipulated data. The processor 504 can be any type of processor, multiprocessor, microprocessor, or controller known in the art. The processor 504 is connected to and communicates data with the memory 502 and the interface 506 via an interconnection mechanism (such as a bus or some other data connection). This interconnection mechanism is Figure 5504. In operation, the processor 504 causes data and / or encoded instructions to be read from the non-volatile data storage medium in the memory 502 and written to the high-performance data storage. The processor 504 manipulates the data and / or executes the encoded instructions in the high-performance data storage and copies the manipulated data to the data storage medium after processing is completed.

[0050] Although computing device 500 is shown as an example of a computing device capable of performing the processes disclosed herein, embodiments are not limited to Figure 5 For example, each process may be performed by a computer having Figure 5 The processes disclosed herein may be performed by one or more computing devices having different architectures or components than those shown. For example, a programmable device may include specially programmed dedicated hardware (such as an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a complex programmable logic device (CPLD), and other silicon implementations) or other hardware customized to perform the processes disclosed herein. Thus, the components of a computing device as disclosed herein may be implemented in software, hardware, firmware, or any combination thereof.

[0051] For example, as described above, the processor 504 may be a general purpose processor. However, when executing the instructions described herein (e.g., Figure 2-4 When a specific software process (as depicted in any one of the preceding) is executed, the processor 504 becomes a special-purpose processor capable of performing the following operations: being able to make specific logic-based decisions based on the input data received, and further being able to provide one or more outputs that can be used to control or otherwise inform subsequent processing to be performed by the processor 504 and / or other processors or circuits communicatively coupled to the processor 504. The processor 504 reacts to specific input stimuli in a specific way and generates corresponding outputs based on the input stimuli. In this sense, the structure of the processor 504 according to one embodiment is Figure 2-4 Furthermore, in some example cases, the processor 504 proceeds through a series of logic transitions in which various internal register states and / or other bit cell states, internal or external to the processor 504, may be set to logic high or logic low. This particular sequence of logic transitions is determined by the state of the electrical input signals to the processor 504, and a dedicated structure is provided by the processor 504 in executing Figure 2-4Each software instruction of the process shown is effectively assumed. In particular, those instructions anticipate various stimuli to be received and change the memory state involved accordingly. In this way, the processor 504 can generate and store or otherwise provide useful output signals. Therefore, it can be understood that the processor 504 becomes a special-purpose machine during the execution of the software process, which is capable of processing only specific input signals and presenting specific output signals based on one or more logical operations performed during the execution of each instruction. In at least some examples, the processor 504 is configured to perform a function in which the software is stored in a data storage device (e.g., memory 502) coupled to the processor 504, and the software is configured to cause the processor 504 to proceed with a series of various logical operations that cause the function to be performed.

[0052] Further example embodiments

[0053] The following examples relate to further embodiments, from which numerous permutations and configurations will become apparent.

[0054] Example 1 is a relay node comprising: a memory storing configuration information specifying that the relay node periodically operate in a reduced power mode; a network interface; and at least one processor coupled to the memory and the network interface. The at least one processor is configured to: receive a duty cycle request via the network interface; reconfigure the configuration information to specify that the relay node operate as specified in the duty cycle request; and operate the relay node as specified in the configuration information.

[0055] Example 2 includes the subject matter of Example 1, characterized in that the at least one processor is further configured to: receive sensor data from the sensor node via the network interface; and transmit the sensor data to the gateway.

[0056] Example 3 includes the subject matter of Example 1 or Example 2, wherein the duty cycle request specifies that the relay node operate in full power mode for a duration.

[0057] Example 4 includes the subject matter of Example 3, wherein the at least one processor is further configured to reconfigure the configuration information to specify that the relay node periodically operate in the reduced power mode in response to expiration of the time duration.

[0058] Example 5 is a sensor node comprising: a memory; a network interface; a sensor configured to obtain at least one sensor reading; and at least one processor coupled to the memory and the network interface. The at least one processor is configured to: predict at least one future sensor reading; compare the at least one future sensor reading to at least one predefined criterion; and in response to the at least one future sensor reading satisfying the at least one predefined criterion: generate a duty cycle request; and transmit the duty cycle request to at least one relay node.

[0059] Example 6 includes the subject matter of Example 5, characterized in that the at least one predefined criterion includes a threshold value and the at least one future sensor reading satisfying the at least one predefined criterion includes the at least one future sensor reading exceeding the threshold value.

[0060] Example 7 includes the subject matter of Example 5 or Example 6, characterized in that the at least one future sensor reading includes a plurality of future sensor readings acquired within a predefined time window, the at least one predefined criterion includes a threshold value, and the at least one future sensor reading satisfying the at least one predefined criterion includes each future sensor reading of the plurality of future sensor readings exceeding the threshold value.

[0061] Example 8 includes the subject matter of any of Examples 5-7, characterized in that the at least one future sensor reading includes a plurality of future sensor readings acquired within a predefined time window, the at least one predefined criterion includes a threshold variance, and the at least one future sensor reading satisfying the at least one predefined criterion includes a variance of the plurality of future sensor readings exceeding the threshold variance.

[0062] Example 9 includes the subject matter of any of Examples 5-8, characterized in that the at least one processor is configured to predict the at least one future sensor reading at least in part by executing a model that predicts future sensor readings within a prediction time domain, and is further configured to generate the duty cycle request including a duration equal to the prediction time domain.

[0063] Example 10 includes the subject matter of Example 9, wherein the model comprises one or more of a regression model, a polynomial curve fitting model, a neural network model, and a decision tree model.

[0064] Example 11 includes the subject matter of Example 9 or Example 10, characterized in that the at least one processor is further configured to: receive the at least one sensor reading; and refine the model using the at least one future sensor reading and the at least one sensor reading.

[0065] Example 12 is a sensor network comprising at least one relay node and at least one sensor node. The at least one relay node is configured to: periodically operate in a reduced power mode; receive a duty cycle request; and operate as specified in the duty cycle request. The at least one sensor node is configured to: predict at least one future abnormal sensor reading; generate a duty cycle request in response to predicting the at least one future abnormal sensor reading; and transmit the duty cycle request to the at least one relay node.

[0066] Example 13 includes the subject matter of Example 12, further comprising a gateway, wherein the at least one relay node is further configured to: receive sensor data from the at least one sensor node; and transmit the sensor data to the gateway.

[0067] Example 14 includes the subject matter of Example 12 or Example 13, wherein the duty cycle request specifies that the at least one relay node operate in the full power mode for a duration.

[0068] Example 15 includes the subject matter of Example 14, characterized in that the at least one relay node is further configured to periodically operate in the reduced power mode in response to expiration of the time duration.

[0069] Example 16 includes the subject matter of any of Examples 12-13, characterized in that the at least one sensor node is further configured to predict the at least one future abnormal sensor reading at least in part by executing a model that predicts future sensor readings within a prediction time domain, and is further configured to generate the duty cycle request including a duration equal to the prediction time domain.

[0070] Example 17 includes the subject matter of Example 16, wherein the model comprises one or more of a regression model, a polynomial curve fitting model, a neural network model, and a decision tree model.

[0071] Example 18 includes the subject matter of Example 16 or Example 17, wherein the at least one sensor node is further configured to: obtain at least one sensor reading; and refine the model using the at least one future abnormal sensor reading and the at least one sensor reading.

[0072] Example 19 is a method of controlling power consumed by a sensor network comprising at least one relay node and at least one sensor node. The method comprises the following acts: operating the at least one relay node in a reduced power mode; predicting at least one future abnormal sensor reading at the at least one sensor node; and operating the at least one relay node in a full power mode in response to predicting the at least one future abnormal sensor reading.

[0073] Example 20 includes the subject matter of Example 19, characterized in that operating the at least one relay node in full power mode comprises: generating a duty cycle request in response to predicting the at least one future abnormal sensor reading; transmitting the duty cycle request to the at least one relay node; receiving the duty cycle request by the at least one relay node; and operating the at least one relay node as specified in the duty cycle request.

[0074] Example 21 includes the subject matter of Example 20, characterized in that operating the at least one relay node as specified in the duty cycle request includes operating the at least one relay node in a full power mode for a duration specified in the duty cycle request, and the method further includes operating the at least one relay node in a reduced power mode in response to expiration of the duration.

[0075] Example 22 includes the subject matter of any of Examples 19-21, wherein predicting the at least one future abnormal sensor reading comprises executing a model that predicts future sensor readings within a prediction horizon.

[0076] Example 23 includes the subject matter of Example 22, wherein executing the model comprises executing one or more of a regression model, a polynomial curve fitting model, a neural network model, and a decision tree model.

[0077] Example 24 includes the subject matter of Example 22 or Example 23, further comprising: obtaining at least one sensor reading; and refining the model using the at least one future abnormal sensor reading and the at least one sensor reading.

[0078] Example 25 includes the subject matter of Example 24, wherein obtaining at least one sensor reading comprises obtaining at least one temperature sensor reading.

[0079] Example 26 is a non-transitory computer-readable medium encoded with instructions that, when executed by one or more processors, cause a process for controlling power consumed by a sensor network including a relay node to be implemented. The process includes the following acts: periodically operating the relay node in a reduced power mode; receiving a duty cycle request by the relay node; reconfiguring configuration information stored on the relay node to specify that the relay node operate as specified in the duty cycle request; and operating the relay node as indicated in the configuration information.

[0080] Example 27 includes the subject matter of Example 26, characterized in that the process further comprises: receiving, by the relay node, the sensor data from the sensor node; and transmitting, by the relay node, the sensor data to the gateway.

[0081] Example 28 includes the subject matter of Example 26 or Example 27, wherein the duty cycle request specifies that the relay node operate in full power mode for a duration.

[0082] Example 29 includes the subject matter of Example 28, characterized in that the process further comprises reconfiguring the configuration information to specify that the relay node periodically operate in the reduced power mode in response to expiration of the time duration.

[0083] The terms and expressions used herein are used as descriptive and non-restrictive terms, and when such terms and expressions are used, it is not intended to exclude any equivalents of the features shown and described (or some parts thereof), and it should be recognized that various modifications are possible within the scope of the claims. Accordingly, the claims are intended to cover all such equivalents. Various features, aspects and embodiments are described herein. As will be understood by those skilled in the art, the various features, aspects and embodiments are susceptible to combination with each other and to variations and modifications. The present disclosure should therefore be considered to include such combinations, variations and modifications. The scope of the present disclosure is not limited by this detailed description but by the appended claims. Future applications claiming priority to the present application may claim the disclosed subject matter in different ways and may generally include any collection of one or more limitations as disclosed herein in various ways or otherwise presented.

Claims

1. A relay node, comprising: a memory storing configuration information specifying that the relay node periodically operate in a reduced power mode; Network interface; as well as at least one processor coupled to the memory and the network interface and configured to: receiving a duty cycle request from a sensor node via the network interface; reconfiguring the configuration information to specify that the relay node operate as specified in the duty cycle request, wherein the duty cycle request specifies that the relay node operate in a full power mode for a duration; as well as The relay node operates as specified in the configuration information.

2. The relay node according to claim 1, wherein: The at least one processor is further configured to: receiving sensor data from the sensor node via the network interface; and The sensor data is transmitted to a gateway.

3. The relay node according to claim 1, wherein: The at least one processor is further configured to reconfigure the configuration information to specify that the relay node periodically operate in the power reduced mode in response to expiration of the time duration.

4. A sensor node comprising: Memory; Network interface; a sensor configured to obtain at least one sensor reading; as well as at least one processor coupled to the memory and the network interface and configured to: predicting at least one future sensor reading; comparing the at least one future sensor reading to at least one predefined criterion; as well as In response to the at least one future sensor reading satisfying the at least one predefined criterion: Generate a duty cycle request; as well as The duty cycle request is transmitted to at least one relay node, thereby requesting the at least one relay node to operate as specified in the duty cycle request, wherein the duty cycle request specifies that the at least one relay node operate in a full power mode for a duration.

5. The sensor node according to claim 4, wherein: The at least one predefined criterion includes a threshold value, and the at least one future sensor reading satisfying the at least one predefined criterion includes the at least one future sensor reading exceeding the threshold value.

6. The sensor node according to claim 4, wherein: The at least one future sensor reading includes a plurality of future sensor readings acquired within a predefined time window, the at least one predefined criterion includes a threshold value, and the at least one future sensor reading satisfying the at least one predefined criterion includes each of the plurality of future sensor readings exceeding the threshold value.

7. The sensor node according to claim 4, wherein: The at least one future sensor reading includes a plurality of future sensor readings acquired within a predefined time window, the at least one predefined criterion includes a threshold variance, and the at least one future sensor reading satisfying the at least one predefined criterion includes a variance of the plurality of future sensor readings exceeding the threshold variance.

8. The sensor node according to any one of claims 4 to 7, characterized in that The at least one processor is configured to predict the at least one future sensor reading at least in part by executing a model that predicts future sensor readings within a prediction horizon, and is further configured to generate the duty cycle request including a duration equal to the prediction horizon.

9. The sensor node according to claim 8, wherein: The model includes one or more of a regression model, a polynomial curve fitting model, a neural network model, and a decision tree model.

10. The sensor node according to claim 8, wherein The at least one processor is further configured to: receiving the at least one sensor reading; and The model is refined using the at least one future sensor reading and the at least one sensor reading.

11. A sensor network comprising: At least one relay node, the at least one relay node being configured to: Periodically operating in a reduced power mode; receiving a duty cycle request; as well as operating as specified in the duty cycle request, wherein the duty cycle request specifies that the at least one relay node operate in a full power mode for a duration; as well as At least one sensor node, the at least one sensor node being configured to: predicting at least one future abnormal sensor reading; generating the duty cycle request in response to predicting the at least one future abnormal sensor reading; and The duty cycle request is transmitted to the at least one relay node.

12. The sensor network of claim 11, further comprising a gateway, wherein the at least one relay node is further configured to: receiving sensor data from the at least one sensor node; and The sensor data is transmitted to the gateway.

13. The sensor network according to claim 11, wherein: The at least one relay node is further configured to periodically operate in the power reduced mode in response to expiration of the time duration.

14. The sensor network according to any one of claims 11 to 13, characterized in that The at least one sensor node is further configured to predict the at least one future abnormal sensor reading at least in part by executing a model that predicts future sensor readings within a prediction time horizon, and is further configured to generate the duty cycle request including a duration equal to the prediction time horizon.

15. A method of controlling power consumed by a sensor network comprising at least one relay node and at least one sensor node, the method comprising: operating the at least one relay node in a reduced power mode; predicting at least one future abnormal sensor reading at the at least one sensor node; as well as The at least one relay node is operated in a full power mode in response to predicting the at least one future abnormal sensor reading.

16. The method according to claim 15, wherein Operating the at least one relay node in full power mode comprises: generating a duty cycle request in response to predicting the at least one future abnormal sensor reading; transmitting the duty cycle request to the at least one relay node; receiving, by the at least one relay node, the duty cycle request; and The at least one relay node is operated as specified in the duty cycle request.

17. The method according to claim 16, wherein Operating the at least one relay node as specified in the duty cycle request includes operating the at least one relay node in the full power mode for a duration specified in the duty cycle request, and the method further includes operating the at least one relay node in the reduced power mode in response to expiration of the duration.

18. The method according to any one of claims 15 to 17, wherein Predicting the at least one future abnormal sensor reading includes executing a model that predicts future sensor readings within a prediction horizon.

19. The method of claim 18, further comprising: obtaining at least one sensor reading; as well as The model is refined using the at least one future abnormal sensor reading and the at least one sensor reading.

20. The method according to claim 19, wherein Obtaining the at least one sensor reading includes obtaining at least one temperature sensor reading.

21. A relay node, comprising: means for periodically operating the relay node in a reduced power mode; means for receiving, by the relay node, a duty cycle request from a sensor node; means for reconfiguring configuration information stored on the relay node to specify that the relay node operate as specified in the duty cycle request; as well as Means for operating the relay node as specified in the configuration information, wherein the duty cycle request specifies that the relay node operate in a full power mode for a duration.

22. The relay node of claim 21, further comprising means for reconfiguring the configuration information to specify that the relay node periodically operate in the reduced power mode in response to expiration of the duration.

Images