| CPC G06F 17/11 (2013.01) [G06F 9/505 (2013.01); G06F 9/5072 (2013.01); G06N 7/01 (2023.01); G06Q 10/0631 (2013.01); G06Q 50/40 (2024.01)] | 19 Claims |
|
18. A computing system for incentivizing a plurality of drivers dispersed across a plurality of locations, the computing system comprising:
a processor; and
a non-transitory computer-readable storage medium storing instructions that, when executed by the processor, cause the processor to perform operations comprising:
receiving, via encrypted communication from a client device of each driver of the plurality of drivers, a real-time position of the driver represented by global coordinates measured by a global positioning system (GPS) receiver coupled to the client device of the driver, the real-time positions of the drivers including one or more new drivers;
receiving, via encrypted communication from a client device of each rider of a plurality of riders, a position of the rider represented by global coordinates measured by a GPS received coupled to the client device of the rider;
evaluating an initial set of results for a set of initial candidates according to at least an objective function, the objective function defining a high-dimensional search space and configured to receive as input a plurality of variables related to positioning of drivers and riders dispersed across the plurality of locations and to output a result based on the input, wherein the result relates to an efficiency in completion of one or more tasks by the plurality of drivers at the plurality of locations;
generating a local model based on the initial set of results, by:
generating a Gaussian process posterior distribution based on the initial results of the initial candidates, and
defining a trust region centered around a current best candidate of the initial candidates;
iteratively updating the local model to explore the high-dimensional search space by:
identifying a plurality of new candidates;
determining, for each new candidate of the plurality of new candidates, a prediction of whether the new candidate violates a constraint based on the local model;
determining, for each new candidate of the plurality of new candidates, a utility score for the new candidate, the utility score for the new candidate at least based on the prediction of whether the new candidate violates a constraint;
selecting a new candidate from the plurality of new candidates based on the utility scores of each of the new candidates;
evaluating a subsequent result for the selected new candidate, the subsequent result evaluated according to at least the objective function;
updating the Gaussian process posterior distribution based on the subsequent result;
determining whether the selected new candidate has a utility score greater than a utility score of the current best candidate of the local model; and
responsive to determining that the utility score of the selected new candidate is greater than the utility score of the current best candidate, re-centering the trust region around the selected new candidate;
identifying an optimal solution from the updated local model;
determining an incentive to provide to each driver based on the optimal solution; and
generating, via the client device of each driver, an interactive user interface for each driver displaying a map of a geographic region pertaining to the real-time position of the candidate driver and the incentive, wherein the interactive user interface displays the real-time position of the driver in the map of the geographic region.
|