Computer modeling for discretionary rest predictions

Abstract

Techniques for improving computer modeling for generating predictions are described herein. In one example, a computer model is generated to examine data representing time slots in a schedule. In particular, the computer model can be configured to calculate a homeostatic pressure at a first slot in the schedule and determine (1 ) whether the calculated value of homeostatic pressure at the first slot exceeds a preset threshold and (2) whether the first slot corresponds to a preset status. In response to positive determinations to both, the computer model can iteratively search for a second slot in a portion of the schedule that precedes the first slot such that discretionary operations at the second slot would reduce the homeostatic pressure at the first slot to not exceed the threshold. Thus, the computer model can predict the discretionary operations even without relying on historical records of such operations.

Description

COMPUTER MODELING FOR DISCRETIONARY REST PREDICTIONSCROSS-REFERENCE TO RELATED APPLICATION(S)

[0001] This application claims priority to U.S. Provisional Application No. 63 / 597,476, filed on November 9, 2023, the disclosure of which is incorporated herein by reference.BACKGROUND

[0002] Computer modeling is an important tool that uses heuristics and computational power to simulate, analyze, and predict phenomena or behaviors. At the core, computer modeling involves using computer software to create virtual representations, such as digital models in computer codes that describe target systems or processes. Such digital models can range from mathematical equations to intricate simulations involving multiple variables and interactions in multiple dimensions. By configuring input data and parameters of the digital models, users can simulate or predict how the target systems or processes would behave in response to different conditions or events. As such, computer modeling can allow users to explore scenarios, evaluate hypotheses, predict outcomes, and make informed decisions without costly and timeconsuming physical experiments.SUMMARY

[0003] This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter.

[0004] As discussed above, computer modeling can be an important tool for simulating and analyzing target systems or processes. Typically, computer modeling relies heavily on historical data to function effectively. Historical data serves as a foundation that provides empirical evidence for accurate and dependable predictions from the computer models. For instance, computer models for weather forecasting often draw on decades or even centuries of temperature, precipitation, and atmosphericdata to simulate long-term changes in climate patterns. Economists utilize historical market and financial data to model economic trends, analyze economic cycles, and forecast future economic trends. Similarly, in epidemiology, historical data on disease outbreaks have been used to model spreads of infectious diseases and to evaluate the potential impact of public health interventions.

[0005] A data-driven approach for computer modeling allows for testing of hypotheses, validation of theories, and exploration of scenarios based on historical records. However, in certain situations, historical data may be unavailable, too costly to obtain, or too dynamic to be useful for predicting future behaviors, such as in the field of fatigue risk management. Fatigue from sleep loss or circadian rhythm interruptions can degrade cognitive functioning and necessitates use of countermeasures to maintain performance, productivity, and safety. Errors and accidents resulting from fatigue can cause considerable personal, economic, and societal costs. Fatigue risk management is thus designed for predicting, preventing, and / or mitigating fatigue and related negative consequences.

[0006] A useful tool for fatigue risk management is biomathematical fatigue modeling, i.e., prediction of periods of elevated fatigue levels based on scientific knowledge of biology related to human sleep regulation and circadian rhythm icity. Accurate biomathematical fatigue modeling can be based on certain information, such as when people can sleep and when people would indeed sleep when given an opportunity. In operational settings involving people working long hours and through the night, knowing when people can sleep is usually straightforward (e.g., when off duty, not driving to / from work, and not otherwise required to be awake). Yet, knowing when people would indeed sleep when given an opportunity turns out to be quite challenging especially when people travel across time zones.

[0007] Previous techniques to address such a challenge are mainly based on collecting historical data of when people sleep in particular operational settings. However, such a technique is severely limited by requiring substantial data collection even if such data collection is feasible. There is also no guarantee that findings based on collected historical records in one situation would generalize to new situations in other settings. For example, historical records indicating a certain sleep period for aworker on one working schedule may not accurately predict when the worker would sleep after his / her duty schedule has been changed.

[0008] Several embodiments of the disclosed technology are directed to computer modeling for predicting when discretionary rest would occur without collecting historical records of sleep data. The disclosed technology is based on 1 ) scientific knowledge regarding biology of sleep / wake regulation, including when wakefulness is naturally induced (low probability for sleep), when sleep is naturally induced (high probability for sleep), and when sleep is biologically discretionary; 2) scientific knowledge regarding the level of fatigue that arises when sleep is curtailed or displaced as well as risks to performance, productivity, and safety associated with such fatigue; 3) a realization that biological regulation of sleep is reactive, i.e. , there is natural recovery from curtailed or displaced sleep, but not until after fatigue and the associated risks have already occurred; 4) a recognition that people can, and often would, therefore take discretionary rest (e.g., daytime naps) proactively to mitigate expected future fatigue; and 5) understanding that when sleep is biologically discretionary, proactive rest to mitigate expected future fatigue is also a function of self-selected behavior and decision making.

[0009] As such, based on the foregoing knowledge and understandings, several embodiments of the disclosed technology can provide computer systems and methods configured to determine when discretionary rest would occur without referencing historical records. Without being bound by theory, it is believed that based on guidance from employers, consultants, past personal experience, intuition, or other suitable knowledge or experience, workers can anticipate when future fatigue during an upcoming duty period would be a problem, e.g., exceeding a preset threshold. Likewise, workers have an implicit understanding of when best to proactively rest and for how long to adequately mitigate or preempt the anticipated high levels of future fatigue. Such an understanding is based in the biology of sleep / wake regulation, such that proactive rest is best placed when a probability for sleep is highest during an interval (of, e.g., about one or two days) preceding a duty period because such a rest yields the greatest probability that sleep would occur and be expected to most effectively reduce the anticipated high levels of future fatigue to be below the preset threshold. It is also believed that workers would seek to be efficient in the use of time for such discretionary rest. To that end, a worker would plan a discretionary rest to bejust long enough to adequately mitigate the anticipated high levels of future fatigue but not any longer. As such, with the discretionary rest, the future fatigue can be reduced to be just below the preset threshold, e.g., less than 10%, 5%, 1 %, or any other suitable percentages from the preset threshold.

[0010] Thus, given a duty schedule, estimation of when a worker would proactively rest can be reduced to estimating the optimal or near-optimal placement of one or more discretionary rest periods in a time interval preceding a duty period with anticipated high levels of fatigue without any prior rest. This time interval is referred to herein as a “lookback window.” Example duration of a lookback window can be about one to two days, such as thirty-six hours or any other suitable lengths of time. A lookback window can contain one or more discretionary rest opportunities (i.e. , one or more periods when sleep is not limited by duty periods or constrained by sleep biology). One objective of the disclosed technology is to predict placement of discretionary rest period(s) such that any resulting rest reduces future fatigue during a subsequent duty period to be just below an acceptable level while a probability of the discretionary rest occurring is highest or near-highest and the rest duration is minimal or near-minimal.

[0011] One possible technique for conducting the foregoing estimation is trying every time / duration combination of one or more rest periods within the available rest opportunities to determine which could best accomplish desired future fatigue mitigation. However, the inherent nonlinearities and discontinuities of this estimation would render this technique inefficient and slow, especially when evaluating thousands of duty schedules to predict and address fatigue issues before and during operation. The foregoing estimation technique may also be even unfeasible or impractical because sufficient computational power may be unavailable to generate suitable estimation results in a timely manner.

[0012] To overcome the foregoing constraints, several embodiments of the disclosed technology provide an efficient technique for performing the foregoing estimation. In one example, aspects of the disclosed technology are directed to dividing a duty schedule into “time bins” or “slots” each with a length of 5, 15, 30, or other suitable numbers of minutes, iteratively estimating an optimal or near-optimal timing of discretionary rest at a time bin, assessing the effect of the discretionary rest on the anticipated fatigue during a subsequent duty period, and repeating the foregoingoperations for other time bins until the anticipated fatigue during the duty period has been adequately mitigated (e.g., at least not exceeding the preset threshold).

[0013] It has been recognized that such incrementally placed time bins for discretionary rest tend to cluster into one or more larger consolidated time bins that can adequately mitigate fatigue with the near-highest probability that sleep would occur and with near-minimal investment of time. By tracking locations of the time bins for discretionary rest, a solution to the foregoing estimation can be quickly found. Moreover, at least certain aspects of the disclosed technology can be readily implemented to evaluate an entire duty schedule with multiple duty periods that each may or may not involve proactive rest behavior and can be easily integrated with existing biomathematical fatigue models.

[0014] Several embodiments of the disclosed technology can thus efficiently predict discretionary rest periods of workers to mitigate anticipated future fatigue without collecting historical records of rest patterns. By at least reducing or even avoiding collecting, storing, processing of data of rest patterns, several embodiments of the disclosed technology can achieve the foregoing objective with significantly reduced resource demands in compute, network, storage, and thus significantly reduced power consumption. If historical records happen to be available, several embodiments of the disclosed technology can also utilize the available historical records to augment the expected prediction accuracy. In addition, by utilizing the iterative estimation operations discussed herein, solutions can be found without requiring massive compute power and thus further saving energy consumption and operational time.BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a schematic diagram illustrating a computer-implemented model generator configured to predict proactive discretionary rest without historical records in accordance with embodiments of the disclosed technology.

[0016] Figure 2 is a flow chart illustrating a process of preprocessing data of duty schedule for predicting discretionary rest in accordance with embodiments of the disclosed technology.

[0017] Figure 3 is a flow chart illustrating a process of determining the state of a homeostatic process for predicting discretionary rest in accordance with embodiments of the disclosed technology.

[0018] Figure 4 is a flow chart illustrating a process of iteratively examining time bins of a duty schedule for possible discretionary rest in accordance with embodiments of the disclosed technology.

[0019] Figure 5 is a flow chart illustrating a process of searching for an optimal or near-optimal placement of a discretionary rest period that can efficiently mitigate fatigue during a subsequent duty period if discretionary rest is taken in accordance with embodiments of the disclosed technology.

[0020] Figures 6A-6C are charts of homeostatic pressure versus time-of-day illustrating example homeostatic and circadian processes under various scenarios in accordance with embodiments of the disclosed technology.

[0021] Figure 7 is a computing device suitable for executing at least some of the components of the computer-implemented model generator in Figure 1.DETAILED DESCRIPTION

[0022] Various embodiments of digital data processing facilities, computing systems, computer devices, and related processes of computer modeling for discretionary rest predictions are described herein. In the following description, specific details of examples are included to provide a thorough understanding of certain embodiments of the disclosed technology. A person skilled in the relevant art would also understand that the disclosed technology may have additional embodiments or may be practiced without several of the details of the embodiments described below with reference to Figures 1-7.

[0023] As discussed above, a data-driven approach for computer modeling can allow for the testing of hypotheses, validation of theories, and exploration of scenarios. However, in certain situations, historical data may be unavailable, too costly to obtain, or too dynamic to be useful for predicting future behaviors, such as in the field of fatigue risk management. When sleep is not constrained by work, commuting, caregiver duties, medical conditions, cultural norms, lifestyle, or other commitments, sleep / wakeregulation strongly impacts timing and duration of sleep. For an average healthy adult, the biological processes, i.e., a circadian process that tracks time of day and a homeostatic process that tracks a balance between time spent awake and time spent asleep, drive sleep to occur at night in a single consolidated period approximately eight hours in duration. When nighttime wakefulness for work or other activities interferes with this biologically driven timing and duration of sleep, the biological processes respond reactively, seeking to recover the lost sleep later and ultimately trying to push sleep back into the night. Before sleep is fully recovered, however, there is a period of fatigue because of a) the homeostatic balance between wakefulness and sleep being disturbed (i.e., a state of insufficient sleep); and / or b) wakefulness occurring at night when the circadian process promotes sleep. Such fatigue puts workers at risk of errors and accidents. Thus, anticipating this fatigue is a first step in managing the associated risks.

[0024] One approach to mitigate the foregoing fatigue-related risks is through proactive discretionary rest, e.g., daytime napping before a duty period. Based on guidance from employers, past personal experience, intuition, or other sources, use of this approach is common in around-the-clock operations such as emergency medical services, passenger and cargo flight operations, the military, etc. The effectiveness of proactive discretionary rest depends on the homeostatic and circadian processes of sleep / wake regulation. The homeostatic process builds up a pressure for sleep in a saturating exponential manner when a person is awake. The pressure then dissipates in a saturating exponential manner when a person is asleep. Simultaneously, the circadian process modulates two thresholds in a near-sinusoidal fashion across the twenty-four hours of the day. The upper threshold or “sleep threshold” delineates a sleep-inducing zone. When a person is awake and the homeostatic process rises above the sleep threshold, the person experiences fatigue and naturally would want to sleep unless a person mounts a deliberate effort to stay awake (e.g., to finish a shift). The lower threshold or “wake threshold” delineates a wake-inducing zone. When a person is asleep and the homeostatic process drops below the wake threshold, the person wakes up naturally. If the state of the homeostatic process is between the two thresholds, sleep is biologically neither induced nor avoided. This discretionary zonefor sleep is where people may begin a proactive rest period, or end a sleep period by, for example, setting an alarm.

[0025] Figures 6A-6C show homeostatic pressure versus time-of-day charts illustrating example homeostatic and circadian processes in three different scenarios. In Figures 6A-6C, the solid line represents a homeostatic process, i.e., higher in the graph denoting greater sleep pressure. The dashed line represents an upper threshold modulated by the circadian process to delineate the sleep-inducing zone. The dotted line represents a lower threshold modulated by the circadian process to delineate the wake-inducing zone.

[0026] Figure 6A shows a scenario in which sleep is free of any relevant constraints and naturally occurs at night in a consolidated period of approximately eight hours. As shown in Figure 6A, sleep begins naturally when the homeostatic process that builds up homeostatic pressure during wakefulness exceeds the sleep threshold and reaches the sleep-inducing zone. Similarly, sleep terminates naturally when the homeostatic process that dissipates homeostatic pressure during sleep exceeds the wake threshold and reaches the wake-inducing zone. The net result is that sleep occurs at night in a consolidated period of approximately eight hours.

[0027] Figure 6B shows an example of nighttime sleep being displaced by a nighttime duty period (e.g., a night shift). As shown in Figure 6B, after the homeostatic pressure exceeds the sleep threshold and reaches the sleep inducing zone, no sleep occurs because the worker is still awake. The excess homeostatic pressure within the sleepinducing zone causes fatigue while the worker is awake at night. The excess homeostatic pressure reactively triggers daytime sleep as soon as the worker’s duty period ends to reduce the homeostatic pressure to be below the sleep threshold. Also, the reactive nature of natural sleep regulation indicates that recovery does not occur until after fatigue has already occurred. Thus, sleep is displaced and begins as soon as an opportunity for sleep arises after a duty period. Sleep ends naturally when the homeostatic pressure falls below the wake threshold and reaches the wake-inducing zone even though the duration of sleep is curtailed, and recuperation is incomplete.

[0028] For comparison, Figure 6C shows a proactive placement of a discretionary rest (e.g., a daytime nap) before a duty period to mitigate anticipated high levels of fatigue.The discretionary rest in this scenario is a choice behavior constrained but not dictated by biology. Thus, predicting proactive discretionary rest is not just predicting the dynamics of sleep / wake biology but also behaviors based on discretionary decision making. The inventors have recognized that, based on a biology-influenced predictability, discretionary rest is highly probable for mitigating anticipated high levels of future fatigue. As shown in Figure 6C, a discretionary rest period (e.g., proactive daytime nap between about 19:00 and 21 :00 in the evening) before the night shift can reduce the homeostatic pressure sufficiently to avoid fatigue of a worker during a subsequent night shift. In other words, the proactive daytime nap sufficiently reduced the homeostatic pressure of the worker at the beginning of the duty period such that the homeostatic pressure never exceeds the sleep threshold throughout the entire duty period. In this example, the worker decides not to take another discretionary rest after the night shift, and the next day sleep naturally begins earlier in the evening to compensate for the lost sleep.

[0029] Several embodiments of the disclosed technology can provide prediction of when people would proactively take discretionary rest in around-the-clock operations and other settings without requiring collecting historical records. In particular, several embodiments of the disclosed technology are configured to iteratively estimate an optimal timing of a discretionary rest period (e.g., 5, 15, or 30 minutes), assessing any effect on the anticipated fatigue during a subsequent duty period given discretionary rest, and repeat the foregoing operations for another period until the anticipated fatigue during the duty period has been adequately mitigated. By tracking locations of the discretionary rest periods, a solution to the foregoing estimation can be quickly found. Moreover, at least certain aspects of the disclosed technology can be readily implemented to evaluate an entire duty schedule with multiple duty periods that each may or may not involve proactive rest behavior and can be easily integrated with existing biomathematical fatigue models. In addition, if historical records happen to be available, several embodiments of the disclosed technology can also utilize the available historical records to augment the expected prediction accuracy.

[0030] As such, several embodiments of the disclosed technology can significantly improve biomathematical modeling of fatigue. By at least reducing or even avoiding collecting, storing, processing of any data of historical records, several embodiments ofthe disclosed technology can predict discretionary rest periods with significantly reduced resource demands in compute, network, storage, and thus power consumption. In addition, by utilizing the iterative estimation operations discussed above, solutions can be found without requiring massive compute power and thus further saving energy consumption and operational time.

[0031] Figure 1 is a schematic diagram illustrating a computer-implemented model generator 100 configured to predict proactive discretionary rest without historical records in accordance with embodiments of the disclosed technology. In Figure 1 and in other Figures herein, individual software components, objects, classes, modules, and routines may be a computer program, procedure, or process written as source code in C, C++, C#, Java, and / or other suitable programming languages. A component may include, without limitation, one or more modules, objects, classes, routines, properties, processes, threads, executables, libraries, or other components. Components may be in source or binary form. Components may also include aspects of source code before compilation (e.g., classes, properties, procedures, routines), compiled binary units (e.g., libraries, executables), or artifacts instantiated and used at runtime (e.g., objects, processes, threads).

[0032] Within a system, different components can have different forms. As one example, a system can comprise a first component, a second component, and a third component operatively coupled to one another. The foregoing components can, without limitation, encompass a system that has the first component being a property in source code, the second component being a binary compiled library, and the third component being a thread created at runtime. The computer program, procedure, or process may be compiled into object, intermediate, or machine code and presented for execution by one or more processors of a personal computer, a tablet computer, a network server, a laptop computer, a smartphone, and / or other suitable computing devices.

[0033] Equally, components may include hardware circuitry. In certain examples, hardware may be considered fossilized software, and software may be considered liquefied hardware. As just one example, software instructions in a component may be burned to a Programmable Logic Array circuit or may be designed as a hardware component with appropriate integrated circuits. Equally, hardware may be emulated by -software. Various implementations of source, intermediate, and / or object code and associated data may be stored in a computer memory that includes read-only memory, random-access memory, magnetic disk storage media, optical storage media, flash memory devices, and / or other suitable computer readable storage media. As used herein, the term “computer readable storage media” excludes propagated signals.

[0034] As shown in Figure 1 , the model generator 100 can include a data preprocessor 102, a homeostatic pressure module 104, a prediction engine 104, and a lookback module 108 operatively coupled with one another. Though particular components are shown in Figure 1 , in other embodiments, the model generator 100 may include additional and / or different components in addition to or in lieu of at least one of the foregoing components. For example, in certain implementations, the data preprocessor 102 may be omitted. In other implementations, the homeostatic pressure module 106 may be external and communicatively coupled to the model generator 100. In further implementations, the model generator 100 may include output, network, storage, or other suitable types of component(s).

[0035] The data preprocessor 102 can be configured to receive data representing a duty schedule 101 of one or more workers. The data of duty schedule 101 can include days and time durations during which the one or more workers are on / off duty based on planned or actual work schedules, or work schedules augmented with information about other demands (e.g., preparation time, commutes, training and other non-operational duties, rest requirements / regulations, etc.) and / or on-the-job rest opportunities. Though duty periods are used herein to demonstrate various aspects of the disclosed technology, embodiments of the disclosed techniques can also be applied for other activity periods that may be performance-critical, safety-critical, or otherwise of interest.

[0036] In certain embodiments, the data preprocessor 102 can also be configured to validate the received duty schedule 101 (e.g., by checking for time conflicts) and indicating data errors based on certain preset criteria. In further embodiments, the data preprocessor 102 can be configured to format, segment, initialize, or otherwise prepare the received data of duty schedule 101 for further processing by the prediction engine 104. For example, the data preprocessor can represent the duty schedule 101 in a computer memory by an array of time bins that are marked either “On duty” or “not on duty.” The data preprocessor 102 can also generate another array of time bins to keeptrack of times when sleep is scheduled. Yet another array of time bins can be created to track whether a time bin has been examined for proactive discretionary rest. Further, the data preprocessor 102 can assign an initial value of the homeostatic pressure by computing a state of the homeostatic process from the worker’s sleep / wake history or by making a reasonable assumption (e.g., that the worker is well-rested at the start of the duty schedule). One example data preparation process 110 performable by the data preprocessor 102 is discussed below with reference to Figure 2.

[0037] The homeostatic pressure module 104 can be configured to determine a state of the homeostatic process for a worker for a particular time bin. In certain embodiments, the state of the homeostatic process can be represented by a homeostatic pressure or a fatigue level based on a circadian process of the worker. In other embodiments, the state of the homeostatic can be represented by a combination of the foregoing parameters and / or other suitable parameters. In one implementation, the homeostatic pressure module 104 can be configured to calculate the homeostatic pressure or the fatigue level based on the following equations for the homeostatic process (function S) and the circadian process (function C) as follows:where t represents the time associated with a time bin and At is the time step; Tr and Td are the rise (increase) and decline (decrease) time constants of the change in homeostatic pressure during wakefulness and sleep, respectively; and A, to and T are the amplitude, phase, and period of the circadian process, respectively (where typically T = 24 hours). In other implementations, alternative sets of equations and parameter values can also be suitable for determining the state of the homeostatic process. Anexample process 150 for determining the state of homeostatic process performable by the data preprocessor 102 is discussed below with reference to Figure 3.

[0038] The prediction engine 106 can be configured to iteratively examine time bins of the received duty schedule 101 for possible discretionary rest. In certain embodiments, the prediction engine 106 can start a loop for examining a time bin or slot at an initial time ti, which can be a user selected beginning time bin in the duty schedule 101 , a beginning duty period in the duty schedule 101 , or other suitable times. The prediction engine 106 can then be configured to determine whether (1 ) a subsequent time bin ti+i at falls at least in part in a duty period according to the duty schedule 101 ; and (2) a fatigue level in the time bin ti+i exceeds a preset fatigue threshold. In response to determining that both of the foregoing conditions are satisfied, the prediction engine 106 can be configured to invoke the lookback module 108 that can iteratively search for an optimal or near-optimal placement of a period prior to the duty period at ti+i such that discretionary rest during the period mitigates fatigue during the duty period.

[0039] In response to determining that the time bin is not a duty period, the prediction engine 106 can be configured to assign a state of “Sleep,” “Wake,” or maintain a prior state based on a homeostatic pressure level for the time bin at ti+i . After examining the time bin at ti, the prediction engine 106 can be configured to examine another time bin in a similar fashion by, for example, incrementing a slot count by a preset quantity, e.g., by one. Thus, iterating through all the time bins in the duty schedule 101 , the prediction engine 106 can generate and output a prediction result 103 that shows not only reactive but also proactive discretionary rest period(s) of the worker based on both the homeostatic and circadian processes. An example process 120 for iteratively examining time bins for possible discretionary rest is discussed in more detail below with reference to Figure 4.

[0040] The lookback module 108 can be configured to search for an optimal or near- optimal placement of a discretionary rest period that can efficiently mitigate fatigue during a subsequent duty period if discretionary rest is taken. In certain embodiments, the lookback module 108 can be configured to isolate a lookback window as a set of time bins immediately prior to ti+i. Within the defined lookback window, the lookback module 108 can be configured to generate a lookback subset of time bins that are (1 ) not in a “Sleep” state; (2) not in a “On duty” state; and (3) not having a flag indicatingthe time bin(s) have been examined. The lookback module 108 can then determine whether the lookback subset includes any time bins at all. If no time bins are present in the lookback subset, the lookback module 108 can be configured to terminate execution and indicate to the prediction engine 106 that no solution can be found. If at least one time bin, e.g., at time tj is present in the lookback subset, the lookback module 108 can be configured to locate a time bin at tk at which the worker has the highest fatigue level and assign a “Sleep” state to the time bin at tk. In one implementation, the fatigue level can be defined as a homeostatic pressure relative to the lower threshold. In other implementations, the fatigue level may be defined in other suitable manners.

[0041] The lookback module 108 can then be configured to determine whether setting a “Sleep” state to the time bin at tk would cause any time bins with a “Sleep” state in the lookback window to have a homeostatic pressure less than the wake threshold. If at least one time bin with a “Sleep” state in the lookback window has a homeostatic pressure less than the wake threshold, the lookback module 108 can be configured to reassign a “Wake” state to the time bin at tk and mark the time bin at tk as examined. Otherwise, the lookback module 108 maintains the assigned “Sleep” state at the time bin tk and mark the time bin as examined. The lookback module 108 can then be configured to determine whether the fatigue level at the time bin at ti+i is still above the fatigue threshold. If not, the lookback module 108 returns results to the prediction module 106. Otherwise, the lookback module 108 reverts to defining another lookback subset in an iterative manner until the fatigue level of the time bin at ti+i is no longer above the fatigue threshold, or all time bins in the lookback window has been examined, whichever comes first. An example process 160 of searching for an optimal or near-optimal placement of a discretionary rest period that can efficiently mitigate fatigue during a subsequent duty period if discretionary rest is taken is discussed below with reference to Figure 5.

[0042] Figure 2 is a flow chart illustrating a process 110 of preprocessing data of duty schedule 101 (Figure 1) for predicting discretionary rest in accordance with embodiments of the disclosed technology. In the illustrated embodiment, the data preprocessor 102 (Figure 2) can be configured to execute some or all of the stages of process 110. In other embodiments, other suitable component(s) of the modelgenerator 100 (Figure 2) can be configured to execute at least one of the stages of process 110 described below.

[0043] As shown in Figure 2, the process 110 can include an optional stage 111 for validating the received data of duty schedule 101 (Figure 1 ). In certain embodiments, the duty schedule 101 can be analyzed for time conflicts; for example, a time period may be marked as both on duty and off duty. In other embodiments, validating the received data can include checking for authenticity (e.g., based on digital certificates / signatures) or indicating data errors based on certain preset criteria. In further embodiments, the validating stage 110 may be omitted from the process 110.

[0044] The process 110 can also include creating one or more arrays of time bins or slots based on the received duty schedule at stage 112. In the embodiment shown in Figure 2, a sleep array 113, a duty array 115, and an exam array 117 are shown for illustration purposes. The sleep array 113 can include columns and rows of data fields representing, for instance, a start time of a time bin, an end time of a time bin, and a data field containing data indicating a state of “Sleep” or “Wake.” Similarly, the duty array 115 can include columns and rows of data fields representing, for instance, a start time of a time bin, an end time of a time bin, and a data field containing data indicating a state of “On duty” or “Off duty.” In addition, the exam array 117 can include columns and rows of data fields representing, for instance, a start time of a time bin, an end time of a time bin, and a data field containing data indicating whether a time bin has been “Examined” or “Not examined.” In other embodiments, the foregoing arrays may include additional and / or different data fields, such as those containing data indicating a sequence number, count, or other suitable information. In further embodiments, arrays, tables, or other suitable data structures in addition to or in lieu of the foregoing arrays can also be created to facilitate the operations of the prediction module 106 (Figure 1 ).

[0045] The process 110 can then include one or more stages to initialize certain data fields in the foregoing arrays. For example, in the embodiment shown in Figure 2, the process 110 can include setting all slots in the sleep array 113 to “Wake” at stage 114 and setting all slots in the exam array 117 to “Not examined” at stage 116 in parallel. In other embodiments, initialization of the data fields in the foregoing arrays can be performed in sequence, staggered, or in other suitable manners. After initializing the arrays, the operation of the process 110 returns.

[0046] Figure 3 is a flow chart illustrating a process 150 of determining the state of a homeostatic process for predicting discretionary rest in accordance with embodiments of the disclosed technology. In the illustrated embodiment, the homeostatic pressure module 104 (Figure 1 ) can be configured to execute some or all of the stages of process 150. In other embodiments, other suitable component(s) of the model generator 100 (Figure 1 ) can be configured to execute at least one of the stages of process 150 described below.

[0047] As shown in Figure 3, the process 150 can include a decision stage 152 to determine whether time bin ti is associated with a sleep state. In one embodiment, the foregoing determination can be based on data contained in a data field in the sleep array 113 (Figure 2) corresponding to the time bin ti. In other embodiments, the determination can be based on a combination of the data in one or more data fields in the sleep or other arrays. In response to determining that time bin ti is associated with a sleep state, the process 150 can include decreasing a homeostatic pressure level by one increment at stage 154. On the other hand, in response to determining time bin ti is not associated with a sleep state, the process 150 can include increasing a homeostatic pressure level by one increment at stage 156. The increment of homeostatic pressure can be preset, for example, based on a time of day, a duration of the time bin, or other factors. Then, the process 150 includes assigning the derived homeostatic pressure level to a time bin at ti, for example, by modifying data stored in the sleep array 113 (Figure 2) or other suitable data structures. The operation of the process 150 then returns.

[0048] Figure 4 is a flow chart illustrating a process 120 of iteratively examining time bins of a duty schedule for possible discretionary rest in accordance with embodiments of the disclosed technology. In the illustrated embodiment, the prediction engine 106 (Figure 1 ) can be configured to execute some or all of the stages of process 120. In other embodiments, other suitable component(s) of the model generator 100 (Figure 1 ) can be configured to execute at least one of the stages of process 120 described below.

[0049] As shown in Figure 4, the process 150 can include initiating an examination loop for a time bin at t=ti with slot count of i=0 to iteratively examine all slots in the received duty schedule 101 (Figure 1 ). In certain embodiments, the initial time bin canbe a slot at a starting point of the duty schedule 101. In other embodiments, the initial time bin can be a slot at a first duty period in the duty schedule 101 or can be one of other suitable slots in the duty schedule 101. The process 150 then proceeds to determining a homeostatic pressure of a subsequent slot at t=ti+i by, for instance, invoking the homeostatic pressure module 104 (Figure 1). Operations of determining the homeostatic pressure are described in more detail above with reference to Figure 3.

[0050] The process 150 can then proceed to decision stages 124 and 126 to determine whether the subsequent slot at t=ti+i has an “On duty” state and a fatigue level is high in the subsequent slot at t=ti+i . In one implementation, the fatigue level is high if the homeostatic pressure in the subsequent slot at t=ti+i rises above the sleep threshold into the sleep-inducing zone. In other implementations, the fatigue level can be indicated as high based on other suitable criteria. In the illustrated embodiment in Figure 4, the foregoing determination includes a first decision stage 124 to determine whether the slot at t=ti+i has an ”On duty” state by, for instance, looking up data in a corresponding data field in the duty array 115 (Figure 2). In response to determining that the slot at t=ti+i has an ”On duty” state, in one embodiment, the process 150 proceeds to a second decision stage 126 to determine whether fatigue is high for the slot at t=ti+i . In response to determining that fatigue is high for the slot at t=ti+i, the process 150 includes performing a lookback procedure by, for instance, invoking the lookback module 108 (Figure 1 ) to investigate possible mitigation of the detected high fatigue for the slot at t=ti+i via proactive discretionary rest before the slot at t=ti+i. Example operations of performing the lookback procedure are described in more detail below with reference to Figure 5. In other embodiments, the determination at stage 126 can also be based on risks (e.g., probability, magnitude, and / or duration of exposure to one or more adverse consequences associated with fatigue or risks that account for other non-fatigue-related risk factors.

[0051] In response to determining that the slot at t=ti+i does not have an “On duty” state, fatigue level is not high for the slot at t=ti+i , or a lookback procedure has been performed, the process 150 proceeds to assigning either a “Sleep” or “Wake” state to the slot at t=ti+i at stage 129. Though certain decision stages are used for the assignment in Figure 4, in other embodiments, such an assignment can also utilize state machines or other suitable components. As shown in Figure 4, in response todetermining that the slot at t=ti+i does not have an “On duty” state, the process 150 includes another decision stage 134 to determine whether the homeostatic pressure derived at stage 122 at the slot at t=ti+i is low or exceeds the wake threshold. In response to determining that the homeostatic pressure at the slot at t=ti+i is low or exceeds the wake threshold, that fatigue level is not high for the slot at t=ti+i, or the lookback procedure has been invoked, the process 150 proceeds to assigning a “Wake” state by, for instance, setting a data value in the sleep array 113 (Figure 2) corresponding to the slot at t=ti+i .

[0052] In response to determining that the homeostatic pressure at the slot at t=ti+i is not low or does not exceed the wake threshold, the process 150 proceeds to yet another decision stage 136 to determine whether the homeostatic pressure derived at stage 122 is high or exceeds the sleep threshold. In response to determining that the homeostatic pressure derived at stage 122 is high or exceeds the sleep threshold, the process 150 proceeds to assigning a “Sleep” state to the slot at t=ti+i. In response to determining that the homeostatic pressure derived at stage 122 is not high or does not exceed the sleep threshold, the process 150 proceeds to assigning the same “Sleep” or “Wake” state from the slot at t=ti to the slot at t=ti+i .

[0053] Subsequent to assigning a “Sleep” or “Wake” state to the slot at t=ti+i, the process 150 proceeds to another decision stage 142 to determine whether additional slots in the duty schedule 101 are yet to be examined. In response to determining that additional slot(s) exist in the duty schedule 101 , the process 150 proceeds to incrementing the slot count by a step, e.g., one before reverting to determining a homeostatic pressure at stage 122. In response to determining that no additional slots exist in the duty schedule 101 are yet to be examined, the process 150 proceeds to outputting a sleep / wake prediction result at stage 146.

[0054] Figure 5 is a flow chart illustrating a process 160 of searching for an optimal or near-optimal placement of a discretionary rest period in a lookback window that can mitigate high fatigue during a subsequent duty period if discretionary rest is taken in accordance with embodiments of the disclosed technology. The lookback window includes a time interval immediately preceding a time bin at which high fatigue is detected. The lookback window can be twenty four hours, thirty six hours, forty eight hours, or any other suitable durations. In the illustrated embodiment, the lookbackmodule 108 (Figure 1 ) can be configured to execute some or all of the stages of process 160. In other embodiments, other suitable component(s) of the model generator 100 (Figure 1 ) can be configured to execute at least one of the stages of process 160 described below.

[0055] As shown in Figure 5, the process 160 can include identifying a lookback subset of time bins that do not have an “Sleep” state, do not have a “On duty” state, and for which no previous attempt to place a proactive discretionary rest period been made, i.e. , have a “Not examined” state in corresponding data fields of the sleep array 113, the duty array 115, and the exam array 117, respectively, at stage 162. The process 160 can then include a decision stage 163 to determine whether the lookback subset contains any time bins or slots. In response to determining that the lookback subset does not contain any time bins or slots, the process 160 returns. The absence of any time bins in the lookback subset would indicate that previous assessments in the lookback procedure have examined all proactive discretionary rest possibilities, and thus no further proactive fatigue mitigation is possible.

[0056] In response to determining that the lookback subset does contain at least one time bin or slot, the process 160 proceeds to identifying a time bin at t=tk at which the fatigue level is the highest at stage 164. Though many suitable approaches exist for such identification, in one embodiment, a difference between the homeostatic pressure and the sleep threshold (Figures 6A-6C) may be used. The difference provides a measure of excess homeostatic pressure and is an estimate of the level of fatigue. In other embodiments, additional and / or different parameters can be used for the foregoing identification. For example, certain time bins in the lookback window may be assigned greater or less weight to be selected. Including weights can accommodate non-work related constraints or preferences for when workers decide to take proactive discretionary rest. For instance, workers may prioritize sleep during designated times when other activities (e.g., eating, shopping, etc.) may not be possible. People may also use hypnotics or stimulants to promote sleep and wakefulness at specific times.

[0057] The identified time bin at t=tk thus has the highest or near-highest probability that sleep would occur and represents the minimal or near-minimal investment of time for mitigating fatigue. As such, the time bin at t=tk is assumed to reflect an incremental piece of when sleep occurs based on self-selected behaviors and decision making.The time bin at t=tk is therefore assigned a “Sleep” state by, for instance, modifying data in a corresponding data field in the sleep array 113 (Figure 2).

[0058] The foregoing incremental addition of proactive discretionary rest during the lookback window may render it possible that at some point during the lookback window, the homeostatic pressure drops below the lower threshold into the wake-inducing zone. As a result, a previously assigned “Sleep” state is therefore no longer possible as the worker would instead wake up naturally at that time. To address this possibility, the lookback procedure of process 160 includes recomputing (e.g., using a loop) the homeostatic process at stage 166 over the lookback window by, for instance, invoking the process 150 described above with reference to Figure 4. The recomputing can start at the identified time bin at t=tk. The process 160 can then include a decision stage 168 to determine whether the homeostatic pressure drops below the lower or wake threshold anywhere in the lookback window. In response to determining that the homeostatic pressure exceeds the lower threshold for at least one time bin, the process 160 includes indicating that the time bin at t=tk is not a viable option and reassigning a “Wake” state to the time bin at t=tk at stage 170.

[0059] In response to determining that the homeostatic pressure does not exceed the lower threshold anywhere in the lookback window, or a “Wake” state has been reassigned to the time bin at t=tk, the process 160 includes assigning an “Examined” state to the time bin at t=tk at stage 172. The process 160 then proceeds to another decision stage 174 to determine whether the fatigue level at the subsequent time bin at t=ti+i is still high. In response to determining that the fatigue level at the subsequent time bin at t=ti+i is still high, the process 160 reverts to defining another lookback subset at stage 162 that includes time bins that do not have an “Sleep” state, do not have a “On duty” state, and for which no previous attempt to place a proactive discretionary rest period been made. Otherwise, the process 160 returns. As such, with the discretionary rest, the future fatigue can be reduced to a value just below the threshold at the operation at stage 126 in Figure 4, e.g., less than 10%, 5%, 1 %, or any other suitable percentages from the threshold, or at least maximally or near-maximally reduced if the discretionary rest period cannot be further increased in duration given available discretionary rest opportunities.

[0060] Figure 7 is a computing device 300 suitable for executing certain components of the model generator 100 in Figure 1. For example, the computing device 300 can be suitable for the data preprocessor 102, the homeostatic pressure module 104, the predicting engine 106, and / or the lookback module 108 of Figure 1. In a very basic configuration 302, the computing device 300 can include one or more processors 304 and a system memory 306. A memory bus 308 can be used for communicating between processor 304 and system memory 306.

[0061] Depending on the desired configuration, the processor 304 can be of any type including but not limited to a microprocessor (pP), a microcontroller (pC), a digital signal processor (DSP), or any combination thereof. The processor 304 can include one more level of caching, such as a level-one cache 310 and a level-two cache 312, a processor core 314, and registers 316. An example processor core 314 can include an arithmetic logic unit (ALU), a floating-point unit (FPU), a digital signal processing core (DSP Core), a graphical processing unit (GPU), or any combination thereof. An example memory controller 318 can also be used with processor 304, or in some implementations memory controller 318 can be an internal part of processor 304.

[0062] Depending on the desired configuration, the system memory 306 can be of any type including but not limited to volatile memory (such as RAM), non-volatile memory (such as ROM, flash memory, etc.) or any combination thereof. The system memory 306 can include an operating system 320, one or more applications 322, such as the model generator 100 of Figure 2, and program data 324, such as the duty schedule 101 and / or the prediction result 103. This described basic configuration 302 is illustrated in Figure 7 by those components within the inner dashed line.

[0063] The computing device 300 can have additional features or functionality, and additional interfaces to facilitate communications between basic configuration 302 and any other devices and interfaces. For example, a bus / interface controller 330 can be used to facilitate communications between the basic configuration 302 and one or more data storage devices 332 via a storage interface bus 334. The data storage devices 332 can be removable storage devices 336, non-removable storage devices 338, or a combination thereof. Examples of removable storage and non-removable storage devices include magnetic disk devices such as flexible disk drives and hard-disk drives (HDD), optical disk drives such as compact disk (CD) drives or digital versatile disk(DVD) drives, solid state drives (SSD), and tape drives to name a few. Example computer storage media can include volatile and nonvolatile, removable, and nonremovable media implemented in any method or technology for storage of information, such as computer readable instructions, data structures, program modules, or other data. The term “computer readable storage media” or “computer readable storage device” excludes propagated signals and communication media.

[0064] The system memory 306, removable storage devices 336, and non-removable storage devices 338 are examples of computer readable storage media. Computer readable storage media include, but not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other media which can be used to store the desired information, and which can be accessed by computing device 300. Any such computer readable storage media can be a part of computing device 300. In other examples, at least one of the foregoing storage devices may be replaced, supplemented, or otherwise linked to a cloud storage device / service via a computer network.

[0065] The computing device 300 can also include an interface bus 340 for facilitating communication from various interface devices (e.g., output devices 342, peripheral interfaces 344, and communication devices 346) to the basic configuration 302 via bus / interface controller 330. Example output devices 342 include a graphics processing unit 348 and an audio processing unit 350, which can be configured to communicate to various external devices such as a display or speakers via one or more A / V ports 352. Example peripheral interfaces 344 include a serial interface controller 354 or a parallel interface controller 356, which can be configured to communicate with external devices such as input devices (e.g., keyboard, mouse, pen, voice input device, touch input device, etc.) or other peripheral devices (e.g., printer, scanner, etc.) via one or more I / O ports 358. An example communication device 346 includes a network controller 360, which can be arranged to facilitate communications with one or more other computing devices 362 over a network communication link via one or more communication ports 364.

[0066] The network communication link can be one example of a communication media. Communication media can typically be embodied by computer readableinstructions, data structures, program modules, or other data in a modulated data signal, such as a carrier wave or other transport mechanism, and can include any information delivery media. A “modulated data signal” can be a signal that has one or more of its characteristics set or changed in such a manner as to encode information in the signal. By way of example, and not limitation, communication media can include wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, radio frequency (RF), microwave, infrared (IR) and other wireless media. The term computer readable media as used herein can include both storage media and communication media.

[0067] The computing device 300 can be implemented as a portion of a small-form factor portable (or mobile) electronic device such as a cell phone, a personal data assistant (PDA), a personal media player device, a wireless web-watch device, a personal headset device, an application specific device, a smart watch, or a hybrid device that include any of the above functions. The computing device 300 can also be implemented as a personal computer including both laptop computer and non-laptop computer configurations.

[0068] From the foregoing, it would be appreciated that specific embodiments of the technology have been described herein for purposes of illustration, but that various modifications may be made without deviating from the disclosure. In addition, many of the elements of one embodiment may be combined with other embodiments in addition to or in lieu of the elements of the other embodiments. Accordingly, the disclosure is not limited except as by the appended claims.

Claims

CLAIMSI / We claim:

1. A process of computer modeling for discretionary rest predictions, comprising: executing, with a computer processor, instructions to initiate a loop to examine data representing a plurality of sequential slots of time in a schedule, wherein to examine the plurality of sequential time slots includes: calculating a value of homeostatic pressure at a slot in the schedule; determining (1 ) whether the calculated value of homeostatic pressure at the slot exceeds a preset threshold and (2) whether the slot has a preset status associated with being on duty; and in response to determining that the calculated value of homeostatic pressure at the slot exceeds the preset threshold and that the slot has the preset status associated with being on duty, iteratively searching for one or more slots in a portion of the schedule that precedes the slot such that discretionary rest at the one or more slots reduces the homeostatic pressure at the slot to not exceed the preset threshold; and outputting data representing a prediction that discretionary rest would occur at the one or more slots in the schedule without relying on any historical records of such discretionary rest in the schedule.

2. The process of claim 1 wherein calculating the value of homeostatic pressure at the subsequent slot in the schedule includes: determining whether a data field in an array corresponding to a preceding slot contains data indicating sleep during the preceding slot; in response to determining that the data field in the array contains data indicating sleep during the preceding slot, decrementing a value of homeostatic pressure at the preceding slot by a preset amount;in response to determining that the data field in the array contains data indicating being awake during the slot under examination, incrementing the value of homeostatic pressure at the preceding slot by another preset amount; and setting the value of homeostatic pressure at the slot to be the decremented or incremented value of homeostatic pressure.

3. The process of claim 1 wherein the discretionary rest at the one or more slots reduces the homeostatic pressure at the slot to a new value not exceeding the preset threshold and is less than 10%, 5%, or 1 % from the preset threshold.

4. The process of claim 1 wherein the portion of the schedule that precedes the subsequent slot has a preset duration of twenty four, thirty six, or forty eight hours.

5. The process of claim 1 wherein iteratively searching for the one or more slots includes: defining the portion of the schedule that precedes the slot as a lookback window; and identifying a subset of slots in the defined lookback window, the subset of slots individually: having a level of homeostatic pressure not exceeding the preset threshold; not having the preset status; and having not been examined for discretionary rest.

6. The process of claim 1 wherein iteratively searching for the one or more slots includes: defining the portion of the schedule that precedes the slot as a lookback window; identifying a subset of slots in the defined lookback window, the subset of slots individually: having a level of homeostatic pressure not exceeding the preset threshold; not having the preset status; andhaving not been examined for discretionary rest; determining whether the identified subset of slots include at least one slot; and in response to determining that the identified subset of slots includes not one slot, indicating additional discretionary rest is not possible.

7. The process of claim 1 wherein iteratively searching for the one or more slots includes: defining the portion of the schedule that precedes the slot as a lookback window; identifying a subset of slots in the defined lookback window, the subset of slots individually having a level of homeostatic pressure not exceeding the preset threshold, not having the preset status, and having not been examined for discretionary rest; and within the identified subset of slots, selecting a slot corresponding to a level of homeostatic pressure higher than all others in the identified subset of slots; determining whether discretionary rest at the selected slot would reduce homeostatic pressure at the slot under examination to not exceed the preset threshold; and in response to determining that discretionary rest at the selected slot would reduce homeostatic pressure at the subsequent slot to not exceed the preset threshold, generating data representing a prediction that discretionary rest would occur at the selected slot in the identified subset of slots.

8. The process of claim 1 wherein iteratively searching for the one or more slots includes: defining the portion of the schedule that precedes the slot as a lookback window; identifying a subset of slots in the defined lookback window, the subset of slots individually having a level of homeostatic pressure not exceeding the preset threshold, not having the preset status, and having not been examined for discretionary rest; and within the identified subset of slots,selecting a slot corresponding to a level of homeostatic pressure higher than all others in the identified subset of slots; determining whether discretionary rest at the selected slot would reduce homeostatic pressure at the slot under examination to not exceed the preset threshold; and in response to determining that discretionary rest at the selected slot would not reduce homeostatic pressure at the slot under examination to not exceed the preset threshold, marking the selected slot as having been examined; and identifying a new subset of slots in the defined lookback window, the new subset of slots individually having a level of homeostatic pressure not exceeding the preset threshold, not having the preset status, and having not been examined for discretionary rest; and repeating the foregoing selecting and determining operations within the identified new subset of slots.

9. The process of claim 1 wherein iteratively searching for the one or more slots includes: defining the portion of the schedule that precedes the slot as a lookback window; identifying a subset of slots in the defined lookback window, the subset of slots individually having a level of homeostatic pressure not exceeding the preset threshold, not having the preset status, and having not been examined for discretionary rest; and within the identified subset of slots, selecting a slot corresponding to a level of homeostatic pressure higher than all others in the identified subset of slots; determining whether discretionary rest at the selected slot would reduce homeostatic pressure at other slots in the identified subset of slots to be below another preset threshold; and in response to determining that discretionary rest at the selected slot would reduce homeostatic pressure at the other slots in the identified subset of slots to be below the another preset threshold,generating data representing a prediction that discretionary rest would not occur at the selected slot in the subset of slots.

10. A computing device, comprising: a processor; and a memory operatively coupled to the processor, the memory containing instructions executable by the processor to cause the computing device to perform a process according to one of claims 1 -9.

Images