Systems and methods for distributed calculation of fatigue-risk prediction and optimization

• US 8,594,982 B2
• Filed: 06/09/2011
• Issued: 11/26/2013
• Est. Priority Date: 06/09/2011
• Status: Active Grant

• Current Assignee: Pulsar Informatics, Inc.
• Original Assignee: Pulsar Informatics, Inc.
• Inventors: Mott, Christopher Grey, Mollicone, Daniel Joseph, Thomas, Sean Michael, Biondo, Damian Marcus
• Primary Examiner(s): PHAN, THAI Q
• Application Number:
  US13/157,292
• Publication Number:
  US 20120316845A1
• Time in Patent Office: 901 Days
• Field of Search: 703/2, 703/5, 702/95, 702/96, 434/255
• US Class Current: 703/2
• CPC Class Codes: G06F 2209/501 Performance criteria
• CPC Class Codes: G06F 2209/5017 Task decomposition
• CPC Class Codes: G06F 9/5094 where the allocation takes into account power or heat criteria power management in computers in general G06F1/3203; thermal management in computers in general G06F1/206
• CPC Class Codes: G06Q 10/06398 Performance of employee with respect to a job function
• CPC Class Codes: Y02D 10/00 Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

Distributed computing methods and systems are disclosed, wherein intensive fatigue-risk calculations are partitioned according to available computing resources, parameters of the fatigue-risk calculation, time-sensitive user demands, and the like. Methods are disclosed wherein execution-cost functions are used to allocate accessible computing resources. Additional methods include partitioning calculation tasks by user-prioritized needs and by general mathematical features of the calculations themselves. Included herein are methods to calculate only prediction-maximum likelihoods instead of full probability distributions, to calculate prediction likelihoods using Bayesian prediction techniques (instead of full re-tabulation of all data), to collate interim results of fatigue-risk calculations where serial results can be appropriately collated (e.g., serial time-slice independence of the cumulative task involved), to use simplified (e.g., linear, first-order) approximations of richer models of fatigue prediction, to assign user-identified priorities to each computational task within a plurality of such requests, and the like.

Claims

1. A distributed computing method for fulfilling a fatigue-risk calculation task request (“

FPT”

), the method comprising the steps of;

receiving an FPT at a primary computing device,

partitioning the FPT into one or more computational tasks;

calculating, for each of the one or more computational tasks, a set of values for a corresponding set of execution-cost functions, the set of execution-cost functions comprising;

a primary cost function associated with use of the primary computing device to perform the computational task and based at least in part on one or more performance parameters associated with the primary computing device; and

one or more secondary cost functions, each secondary cost function associated with use of a corresponding secondary computing device to perform the computational task and based at least in part on one or more performance parameters associated with the secondary computing device and one or more communication-channel parameters associated with one or more data paths between the secondary computing device and the primary computing device;

allocating, for each of the one or more computational tasks, the computational task in accordance with the set of values for the execution-cost functions associated with the computational task either to;

the primary computing device;

or to an allocated one of the secondary computing devices via an allocated one of the data paths between the allocated secondary computing device and the primary computing device;

transmitting to each of the allocated secondary computing devices the computational tasks allocated thereto respectively via each of the allocated data paths;

calculating a result for each of the one or more computational tasks, wherein for each of the one or more computational tasks, determining the result is performed by the primary computing device if the task has been allocated to the primary computing device and determining the result is otherwise performed by the allocated secondary computing device;

receiving at the primary computing device via the allocated data paths the results of the computational tasks determined by allocated secondary computing devices; and

integrating each of the results of the computational tasks into interim results of increasing completeness until a final solution for the FPT is achieved.

2. The method of claim 1:

wherein the FPT comprises at least in part calculating a fatigue prediction based at least in part on a mathematical fatigue model, and

wherein the step of partitioning the FPT into one or more computational tasks comprises;

creating a computational task to calculate only a prediction maximum likelihood, and

creating a computational task to calculate a full probability distribution prediction.

3. The method of claim 2:

wherein the mathematical model is a state-space model containing at least a prediction function,

wherein the FPT contains at least initial conditions with probability distributions for parameters of the mathematical model,

wherein the prediction maximum likelihood task is calculated using propagation of a maximum likelihood of the initial conditions through the mathematical model'"'"'s prediction function, and

wherein the full probability distribution prediction task is calculated using a Bayesian statistical prediction method.

4. The method of claim 3, wherein the Bayesian statistical prediction method consists essentially of one or more of the following:

a particle filter, an unscented Kalman filter, or a Kalman filter.

5. The method of claim 1:

wherein the FPT consists at least in part of a calculation based at least in part on a mathematical fatigue model, and

wherein the step of partitioning the FPT into one or more computational tasks comprises;

creating a first computational task using a simplified representation of the mathematical fatigue model and

creating a second computational task using the mathematical fatigue model, and

wherein the step of integrating all the results consists of replacing results from the first computational task with results from the second computational task.

6. The method of claim 1:

wherein the FPT comprises a fatigue prediction based at least in part on a mathematical fatigue model that accepts one or more individual profile parameters as input,

wherein the FPT further comprises a task to update the individual profile parameters based on one or more fatigue-related data inputs,

wherein the FPT further comprises current values of the individual profile parameters,

wherein the step of partitioning the FPT into one or more computational tasks consists of;

creating a first computational task to calculate the fatigue prediction using the current values of the individual profile parameters,

creating a second computational task to calculate updated values of the individual profile parameters based on the one or more fatigue-related data inputs, and

creating a third computational task to calculate the fatigue prediction using the updated values of the individual profile parameters, and

wherein the step of integrating all the results consists of replacing results from the first computational task with results from the third computational task.

7. The method of claim 6, wherein the task to update the individual profile parameters consists essentially of using a statistical Bayesian measurement update method.

8. A method of claim 1, wherein the step of partitioning the FPT into one or more computational tasks further comprises:

creating a plurality of computational tasks that produce results having one or more overlapping values, and

assigning a priority value to each computational task within the plurality, and

wherein the step of integrating all the results consists of replacing values of lower priority with overlapping values from results of higher priority.

9. The method of claim 1, wherein the FPT consists of one or more of the following:

a fatigue-level calculation, a fatigue-level calculation over time, a countermeasure recommendation, a countermeasure optimization, a sleep-schedule recommendation, a sleep-schedule optimization, a work-schedule recommendation, or a work-schedule optimization.

10. The method of claim 1, wherein the FPT is related to more than one individual for whom fatigue-related calculations are needed.

11. The method of claim 1, wherein the step of partitioning of the FPT into the one or more computational tasks consists of one or more of the following:

modifying a time resolution associated with the final results of the calculation, modifying the precision or accuracy of the final results, modifying the amount of statistical data included as input to the FPT calculation, using a smaller time window for the FPT calculation, dividing the FPT calculation into multiple time horizons, changing the mathematical model used for fatigue calculations, making a linearizing assumption about the mathematical model selected, calculating only expected values or prediction extremas for future fatigue levels instead of full probability distributions, or conducting the FPT calculation on only a subset of individuals to which the FPT calculation pertains.

12. The method of claim 1, wherein the performance parameters either of the primary computing device or of the one or more secondary computing devices consist essentially of one or more of the following:

processor speed, effective calculation speed, processor power consumption, remaining reserve power available, signal-transmission speed, signal-reception speed, signal-transmission power-consumption rate, or signal-reception power-consumption rate.

13. The method of claim 1, wherein the communication-channel parameters consist essentially of one or more of the following:

data-path send rate, data-path send error rate, effective send rate, send signal power loss, data-path receive rate, data-path receive error rate, effective receive rate, or receive signal power loss.

14. The method of claim 1, wherein the execution cost function can be expressed using one or more of the following terms:

computation time, computation time plus total transfer time, power consumption during calculation, power consumption during transfer, normalized power consumption multiplied by normalized transmission speed, normalized power consumption as limited by maximum time, or a mathematical value reflecting the overall information-security rating of the computing device to which a computational task has been allocated.

15. The method of claim 1, wherein the step of partitioning the FPT into one or more computational tasks and the step of calculating values of execution-cost functions for each of the computational tasks are repeated so as to achieve an optimal partition of the FPT into one or more computational tasks in light of the execution-cost function values.

16. The method of claim 15, wherein the optimal partition of the FPT is determined with respect to user input as to which performance parameters and communication-channel parameters are to be optimized.

17. The method of claim 1, wherein the primary computing device is a mobile device, handheld device, an embedded device, a device containing a central processing unit specifically designed for digital-signal processing, a laptop computer, a notebook computer, a tablet computer, a personal digital assistant, a desktop computer, or any equivalent device.

18. The method of claim 1, further comprising the step of displaying the final solution to the FPT to a user.

19. The method of claim 18, further comprising the step of updating the displaying to a user an interim solution to the FPT upon receipt of each the result of each the computational task until the final solution to the FPT is achieved.

20. The method of claim 1 further comprising the step of updating the performance parameters of the primary computing device or the one or more secondary computing devices regularly or intermittently during operation of the method either through measurement or estimation techniques.

21. The method of claim 1 further comprising the step of transmitting the FPT to a secondary computing device, and wherein the step of partitioning the FPT occurs on the secondary computing device.

22. A distributed computing system for performing a fatigue-risk optimization task (“

FPT”

), the system comprising;

at least two computing devices connected to each other by one or more communications networks, wherein one computing device is designated as a primary computing device and the one or more remaining computing devices are designated as secondary computing devices;

an initializor unit for initializing a plurality of human-fatigue model variables of a dynamic mathematical model that outputs alertness of a human subject based upon said plurality of model variables, said plurality of model variables specifying or estimating probability distributions,

a predictor unit configured to;

estimate current values of said model variables based at least in part on past values of said model variables; and

receive new values for one or more of said model variables when input information is available, said model including a process noise component comprising a probability distribution representing an uncertainty associated with new values for said plurality of model variables;

an alertness estimator unit for using said dynamic mathematical model to estimate alertness values of said human subject based at least in part on said current values of the plurality of model variables;

wherein said predictor unit comprises;

a predictor partitioner subunit to partition one or more of functions of the predictor unit into one or more predictor computational tasks,

a predictor task allocator subunit for allocating each of said one or more predictor computational tasks to a corresponding one of said computing devices,

a predictor task transmission subunit for transmitting one or more of said one or more predictor computational tasks to corresponding ones of said secondary computing devices as allocated by the predictor task allocator via said communications network,

a predictor task-result receiving subunit for receiving results of said predictor computational tasks performed by said computing devices, and

a predictor result integrator subunit for properly combining said results of said predictor computational tasks received at said predictor task-result receiving subunit; and

wherein said alertness estimator unit comprises;

an alertness estimator partitioner subunit to partition one or more of functions of the alertness estimator unit into one or more alertness estimator computational tasks,

an alertness estimator task allocator subunit for allocating each of said one or more alertness estimator computational tasks to a corresponding one of said computing devices,

an alertness estimator task transmission subunit for transmitting one or more of said one or more alertness estimator computational tasks to corresponding ones of said secondary computing devices as allocated by the alertness estimator task allocator via said communications network,

an alertness estimator task-result receiving subunit for receiving results of said alertness estimator computational tasks performed by said computing devices, and

an alertness estimator result integrator subunit for properly combining said results of said alertness estimator computational tasks received at said alertness estimator task-result receiving subunit.

23. A distributed computing system for performing a fatigue-risk optimization task (“

FPT”

), the system comprising;

a plurality of computing devices connected to each other by one or more communications networks, the plurality of computing devices comprising one primary computing device and one or more secondary computing devices;

wherein each of the plurality of computing devices comprises one or more processors configured to provide;

an initializor unit for initializing a plurality of human-fatigue model variables of a dynamic mathematical model that outputs alertness data of a human subject based upon the plurality of model variables, the plurality of model variables specifying or estimating probability distributions,

a predictor unit configured to;

estimate current values of the model variables based at least in part on past values of said model variables;

receive new values for one or more of the model variables when input information is available, the model including a process noise component comprising a probability distribution representing an uncertainty associated with new values for the plurality of model variables; and

an alertness estimator unit for using the dynamic mathematical model to estimate alertness values of the human subject based at least in part on the current values of the plurality of model variables;

wherein the primary computing device further comprises one or more processors configured to provide;

a global FPT process distributor, the global FPT process distributor comprising;

a partitioner subunit for partitioning the FPT into one or more computational tasks,

a task allocator subunit for allocating each of the one or more computational tasks to a corresponding one of the primary or secondary computing devices,

a task transmitter subunit for transmitting one or more of the one or more computational tasks to corresponding ones of the secondary computing devices as allocated by the task allocator subunit via the communications network,

a task result receiver subunit for receiving results of the computational tasks performed by the primary and secondary computing devices, and

a result integrator subunit for properly combining the results of the computational tasks received at the task-result receiver subunit.

24. A computer readable medium carrying instructions that when executed by a suitably configured processor cause the processor to perform the method comprising the steps of:

receiving an FPT at a primary computing device,

partitioning the FPT into one or more computational tasks;

calculating, for each of the one or more computational tasks, a set of values for a corresponding set of execution-cost functions, the set of execution-cost functions comprising;

a primary cost function associated with use of the primary computing device to perform the computational task and based at least in part on one or more performance parameters associated with the primary computing device; and

one or more secondary cost functions, each secondary cost function associated with use of a corresponding secondary computing device to perform the computational task and based at least in part on one or more performance parameters associated with the secondary computing device and one or more communication-channel parameters associated with one or more data paths between the secondary computing device and the primary computing device;

allocating, for each of the one or more computational tasks, the computational task in accordance with the set of values for the execution-cost functions associated with the computational task either to;

the primary computing device;

or to an allocated one of the secondary computing devices via an allocated one of the data paths between the allocated secondary computing device and the primary computing device;

transmitting to each of the allocated secondary computing devices the computational tasks allocated thereto respectively via each of the allocated data paths;

calculating a result for each of the one or more computational tasks, wherein for each of the one or more computational tasks, determining the result is performed by the primary computing device if the task has been allocated to the primary computing device and determining the result is otherwise performed by the allocated secondary computing device;

receiving at the primary computing device via the allocated data paths the results of the computational tasks determined by allocated secondary computing devices; and

integrating each of the results of the computational tasks into interim results of increasing completeness until a final solution for the FPT is achieved.