Predictive utilization of physical resources

1 . A computer-implemented method, comprising:

receiving positioning data associated with a plurality of transportation resources, the positioning data including speed data, location data, and timestamp information for the plurality of transportation resources;

receiving geographic data associated with a geographic region, the geographic data being a graph with nodes representing geographic locations in the geographic region and edges representing road segments connecting the geographic locations;

generating, based at least in part on the positioning data, a first set of embeddings representing temporal information of the positioning data;

generating, based at least in part on the geographic data, a second set of embeddings representing road segment information of individual road segments of the road segments;

generating, based at least in part on the geographic data, a third set of embeddings representing attribute information of the individual road segments;

generating a first input to a first machine learning model based at least in part on the first set of embeddings, the second set of embeddings, and the third set of embeddings;

receiving a first output of the first machine learning model based at least in part on the first input, the first output indicating a first predicted speed value associated with a road segment of the road segments for a transportation resource;

generating a second input to a second machine learning model based at least in part on the first input, the first predicted speed value, and one or more properties associated with a transportation load; and

receiving a second output of the second machine learning model based at least in part on the second input, the second output indicating a distribution of a time length for transporting the transportation load using the transportation resource, wherein the second output is usable in deploying the transportation resource for transporting the transportation load.

2 . The computer-implemented method of claim 1 , wherein training the first machine learning model comprises:

receiving an observed speed value for the road segment of the road segments;

receiving, by the first machine learning model and for the road segment, a training road segment information embedding, a training temporal information embedding, and a training attribute embedding;

receiving, from the first machine learning model, a second predicted speed value for the road segment; and

training the first machine learning model to minimize an error between the observed speed value and the second predicted speed value.

3 . The computer-implemented method of claim 2 , wherein the first predicted speed value, the second predicted speed value, and the observed speed value are normalized with a speed limit of the road segment.

4 . The computer-implemented method of claim 1 , further comprising:

generating a mapping between the positioning data and the road segments based at least in part on the location data; and

generating the second set of embeddings and the third set of embeddings further based at least in part on the mapping.

5 . One or more non-transitory computer-readable media comprising computer-executable instructions that, when executed by one or more computer systems, cause the one or more computer systems to perform operations comprising:

generating a first input to a first machine learning model based at least in part on temporal information, road segment information, and attribute information for a plurality of road segments in a geographic region;

receiving a first output of the first machine learning model based at least in part on the first input, the first output indicating a predicted speed value associated with a road segment of the plurality of road segments for a transportation resource;

generating a second input to a second machine learning model based at least in part on the first input, the predicted speed value, and one or more properties associated with a transportation load; and

receiving a second output of the second machine learning model based at least in part on the second input, the second output indicating a distribution of a time length for transporting the transportation load using the transportation resource, wherein the second output is usable in deploying the transportation resource for transporting the transportation load.

6 . The one or more non-transitory computer-readable media of claim 5 , further comprising additional computer-executable that instructions that, when executed by the one or more computer systems, cause the one or more computer systems to perform additional operations comprising:

receiving positioning data associated with a plurality of transportation resources, the positioning data including speed data, location data, and timestamp information for the plurality of transportation resources; and

receiving geographic data associated with the geographic region, the geographic data being a graph with nodes representing geographic locations in the geographic region and edges representing road segments connecting the geographic locations.

7 . The one or more non-transitory computer-readable media of claim 6 , further comprising additional computer-executable that instructions that, when executed by the one or more computer systems, cause the one or more computer systems to perform additional operations comprising:

generating, based at least in part on the positioning data, a first set of embeddings representing the temporal information of the positioning data;

generating, based at least in part on the geographic data, a second set of embeddings representing the road segment information of individual road segments of the plurality of road segments; and

generating, based at least in part on the geographic data, a third set of embeddings representing the attribute information of the individual road segments.

8 . The one or more non-transitory computer-readable media of claim 5 , wherein, while generating the first output, the first machine learning model is configured to increase a weight of at least one of the temporal information, the road segment information, or the attribute information for a set of road segments of the plurality of road segments that are closer to the road segment at a time of a year in an embedding space.

9 . The one or more non-transitory computer-readable media of claim 5 , wherein the temporal information represents an hour of a day of a week within a year.

10 . The one or more non-transitory computer-readable media of claim 5 , wherein the one or more properties associated with the transportation load comprise at least one of a weight of the transportation load, a type of the transportation resource, or a type of the transportation load.

11 . The one or more non-transitory computer-readable media of claim 5 , further comprising additional computer-executable that instructions that, when executed by the one or more computer systems, cause the one or more computer systems to perform additional operations comprising training the second machine learning model by:

determining distribution parameters for a predicted distribution;

determining the predicted distribution based at least in part on the distribution parameters; and

minimizing a difference between the predicted distribution and an observed time length for transporting a load using a negative log-likelihood function.

12 . The one or more non-transitory computer-readable media of claim 5 , wherein the second output corresponds to a percentile of the distribution, and wherein the percentile is selected based at least in part on a priority of the transportation load.

13 . A system, comprising:

one or more memories configured to store computer-executable instructions;

one or more processors configured to access the one or more memories and execute the computer-executable instructions to at least:

generate a first input to a first machine learning model based at least in part on temporal information, road segment information, and attribute information for a plurality of road segments in a geographic region;

receive a first output of the first machine learning model based at least in part on the first input, the first output indicating a predicted speed value associated with a road segment of the plurality of road segments for a transportation resource;

generate a second input to a second machine learning model based at least in part on the first input, the predicted speed value, and one or more properties associated with a transportation load; and

receive a second output of the second machine learning model based at least in part on the second input, the second output indicating a distribution of a time length for transporting the transportation load using the transportation resource, wherein the second output is usable in deploying the transportation resource for transporting the transportation load.

14 . The system of claim 13 , wherein the one or more processors are configured to access the one or more memories and execute additional computer-executable instructions to at least:

receive positioning data associated with a plurality of transportation resources, the positioning data including speed data, location data, and timestamp information for the plurality of transportation resources; and

receive geographic data associated with the geographic region, the geographic data being a graph with nodes representing geographic locations in the geographic region and edges representing road segments connecting the geographic locations.

15 . The system of claim 14 , wherein the one or more processors are configured to access the one or more memories and execute additional computer-executable instructions to at least:

generate, based at least in part on the positioning data, a first set of embeddings representing the temporal information of the positioning data;

generate, based at least in part on the geographic data, a second set of embeddings representing the road segment information of individual road segments of the plurality of road segments; and

generate, based at least in part on the geographic data, a third set of embeddings representing the attribute information of the individual road segments.

16 . The system of claim 13 , wherein the geographic region is associated with a route for transporting the transportation load, and wherein the second output indicates the distribution of the time length for the route.

17 . The system of claim 13 , wherein the first input comprises a first set of embeddings representing the temporal information, a second set of embeddings representing the road segment information, and a third set of embeddings representing the attribute information.

18 . The system of claim 13 , wherein the distribution is associated with an hour an hour of a day of a week within a year.

19 . The system of claim 13 , wherein the one or more processors are configured to access the one or more memories and execute additional computer-executable instructions to at least:

send a message to a deployment system for deploying the transportation resource, wherein the message includes information for the deployment system to deploy the transportation resource for transporting the transportation load, and wherein the message identifies the transportation load and the transportation resource.

20 . The system of claim 13 , wherein the one or more processors are configured to access the one or more memories and execute additional computer-executable instructions to at least:

generate a mapping between positioning data associated with a plurality of transportation resources and the plurality of road segments based at least in part on location data of the positioning data; and

generate the road segment information and the attribute information based at least in part on the mapping.