| CPC G06F 17/16 (2013.01) [G06F 17/18 (2013.01); G06N 5/045 (2013.01)] | 10 Claims |
|
1. A method of sensor processing, the method comprising:
receiving metadata from a group of physical sensors, wherein
the metadata comprises multiple characteristics of each respective physical sensor of the group of physical sensors, and
the group of physical sensors forms a sensor network;
computing a plurality of different possible subgroups of physical sensors from the group of physical sensors;
in response to detecting that a utility function to select preferable sensors from the group of physical sensors is at least partially unknown, computing a plurality of utility functions by a pseudorandom utility function generator, wherein
each utility function of the plurality of utility functions is monotonically nondecreasing with respect to a dominance relation,
the pseudorandom utility function generator selects at least one random function from a weighted function, a weighted polynomial function, or a weighted power function, and for each utility function of the plurality of utility functions:
computing a utility value for each of the plurality of different possible subgroups of physical sensors based on utility values of the multiple characteristics of the each respective physical sensor, and
identifying a first subset of multiple ones of the plurality of different possible subgroups of physical sensors, wherein the multiple ones of the plurality of different possible subgroups of physical sensors maximize the computed utility value;
controlling execution of a greedy algorithm to iteratively add at least one of the plurality of different possible subgroups of physical sensors of the first subset to at least one cross-function subgroup of physical sensors until a tolerance is reached;
selecting, based on the control of the execution of the greedy algorithm, the at least one cross-function subgroup of physical sensors associated with a cross-function decision option that maximizes multiple computed utility values from the plurality of different possible subgroups of physical sensors of the first subset, wherein
respective ones of the multiple computed utility values are directed to a distinct utility function of the plurality of utility functions,
the at least one cross-function subgroup of physical sensors is included in a maximum number of the plurality of different possible subgroups of physical sensors of the first subset, and the multiple computed utility values include the computed utility value;
selecting one subgroup of the plurality of different possible subgroups of physical sensors, based on the selected at least one cross-function subgroup of physical sensors;
collecting sensory data only from the physical sensors in the selected one subgroup, and combining the collected sensory data into a combined sensory datum;
controlling, based on the combined sensory datum, the selected one subgroup of physical sensors in the sensor network to at least one of control a vehicle, detect a security condition in a security network, detect a health condition, or detect an anomaly in a specific system;
repeating each of the reception of the metadata, the computation of the plurality of different possible subgroups of physical sensors, the computation of the plurality of utility functions, the selection of the at least one cross-function subgroup of physical sensors, the selection of the one subgroup, the collection of the sensory data, and the control of the selected one subgroup for different points in time based on a change in at least one of the multiple characteristics of each of the group of physical sensors over time, wherein
the change is based on environmental conditions associated with the group of physical sensors being different at the different points in time, such that a different subgroup of the different possible subgroups of physical sensors is selected in the repetition at the different points in time; and
for each of the different points in time, execute a calculation at each respective point in time based on the combined sensory datum.
|