V2X message processing for machine learning applications
Methods and systems for processing vehicle to everything (V2X) messages for use by machine learning applications are disclosed. From each of a plurality of vehicles, one or more V2X messages are received, each V2X message including vehicle-related data associated with the vehicle and the received message. A sequence of frames is generated based on the vehicle-related data from at least a subset of vehicles in the plurality of vehicles. Slices of the sequence of frames are aggregated to generate a plurality of time-lapse images. One or more time-lapse images are processed using a machine learning algorithm to generate an output indicative of a traffic-related prediction.
1. A method for processing vehicle to everything (V2X) messages for use by machine learning applications, comprising, by a processor:
receiving, from each of a plurality of vehicles, one or more V2X messages, each V2X message including vehicle-related data associated with the vehicle and the received message, the vehicle-related data indicative at least of a vehicle identifier, message time, and vehicle current location; and
generating, based on the vehicle-related data from at least a subset of vehicles in the plurality of vehicles, a sequence of frames indicative of the locations over time of each vehicle in the subset of vehicles.
2. The method of claim 1 , further comprising, by the processor:
aggregating slices of the sequence of frames to generate a plurality of time-lapse images, each time-lapse image indicative of an approximate location at a given time of each vehicle in the subset of vehicles.
3. The method of claim 2 , wherein the generating a sequence of frames comprises:
for each message in a subset of received messages, generating a respective Boolean grid in which each cell in the grid corresponds to a geographical location, and associating the grid with respective vehicle-related data received in the message;
for each grid, marking, based on the associated vehicle-related data, each cell that corresponds to a location occupied by the vehicle originating the message when the message is transmitted; and
aggregating a plurality of grids in temporal order to generate the sequence of frames.
4. The method of claim 1 , further comprising, by the processor:
processing one or more time-lapse images using a machine learning algorithm to generate an output indicative of a traffic-related prediction.
5. The method of claim 4 , wherein the machine learning algorithm is a neural network machine learning algorithm.
6. The method of claim 4 , wherein the one or more time-lapse images are each associated with a given intersection, and wherein the traffic-related prediction associates the given intersection with a given intersection type from one of several predetermined intersection types.
7. The method of claim 6 , wherein the given intersection type is selected from the group consisting of a bridge, a roundabout and a four-way cross.
8. The method of claim 4 , wherein the generating a sequence of frames comprises:
for each message in a subset of received messages, generating a respective Boolean grid in which each cell in the grid corresponds to a geographical location, and associating the grid with respective vehicle-related data received in the message;
for each grid, marking, based on the associated vehicle-related data, each cell that corresponds to a location occupied by the vehicle originating the message when the message is transmitted; and
aggregating a plurality of grids in temporal order to generate the sequence of frames.
9. The method of claim 1 , wherein the generating a sequence of frames comprises:
for each message in a subset of received messages, generating a respective Boolean grid in which each cell in the grid corresponds to a geographical location, and associating the grid with respective vehicle-related data received in the message;
for each grid, marking, based on the associated vehicle-related data, each cell that corresponds to a location occupied by the vehicle originating the message when the message is transmitted; and
aggregating a plurality of grids in temporal order to generate the sequence of frames.
10. The method of claim 1 , wherein the vehicle-related data further comprises data indicative of at least one of the vehicle's speed and heading.
11. A system for processing V2X messages for use by machine learning applications, comprising:
a message collection module configured to receive, from each of a plurality of vehicles, one or more V2X messages, each V2X message including vehicle-related data associated with the vehicle and the message, the vehicle-related data indicative at least of a vehicle identifier, message time, and vehicle current location; and
a data generation module configured to generate, based on the vehicle-related data from at least a subset of vehicles in the plurality of vehicles, a sequence of frames indicative of the locations over time of each vehicle in the subset of vehicles.
12. The system of claim 11 , wherein the data generation module is further configured to aggregate slices of the sequence of frames to generate a plurality of time-lapse images, each time-lapse image indicative of an approximate location at a given time of each vehicle in the subset of vehicles.
13. The system of claim 12 , further comprising a machine learning module configured to classify the plurality of time-lapse images using a machine learning algorithm and generate an output indicative of a traffic-related prediction.
14. The system of claim 13 , wherein the machine learning algorithm is a neural network machine learning algorithm.
15. The system of claim 13 , wherein the one or more time-lapse images are each associated with a given intersection, and wherein the traffic-related prediction associates the given intersection with a given intersection type from one of several predetermined intersection types.
16. The system of claim 15 , wherein the given intersection type is selected from the group consisting of a bridge, a roundabout and a four-way cross.
17. The system of claim 13 , wherein the data generation module is further configured to:
for each message in a subset of the received messages, generate a respective Boolean grid in which each cell in the grid corresponds to a geographical location and associating the grid with the respective message's vehicle-related data;
for each grid, mark, based on the associated vehicle-related data, each cell that corresponds to a location occupied by the vehicle originating the message at the time of the message; and
aggregate a plurality of grids in temporal order to generate the sequence of frames.
18. The system of claim 12 , wherein the data generation module is further configured to:
for each message in a subset of the received messages, generate a respective Boolean grid in which each cell in the grid corresponds to a geographical location and associating the grid with the respective message's vehicle-related data;
for each grid, mark, based on the associated vehicle-related data, each cell that corresponds to a location occupied by the vehicle originating the message at the time of the message; and
aggregate a plurality of grids in temporal order to generate the sequence of frames.
19. The system of claim 11 , wherein the data generation module is further configured to:
for each message in a subset of the received messages, generate a respective Boolean grid in which each cell in the grid corresponds to a geographical location and associating the grid with the respective message's vehicle-related data;
for each grid, mark, based on the associated vehicle-related data, each cell that corresponds to a location occupied by the vehicle originating the message at the time of the message; and
aggregate a plurality of grids in temporal order to generate the sequence of frames.
20. The system of claim 11 , wherein the vehicle-related data further comprises data indicative of at least one of the vehicle's speed and heading.