Telemetry reporting in vehicle super resolution systems

1. A method comprising:

detecting, by a processor of a vehicle, a difference between a physical characteristic of the vehicle predicted by a first machine learning-based model and a physical characteristic of the vehicle indicated by telemetry data generated by a sub-system of the vehicle;

forming, by the processor, a packet payload of an update packet indicative of the detected difference, based in part on a relevancy of the physical characteristic to the first machine learning-based model;

applying, by the processor, a synchronization strategy to the update packet, to synchronize the update packet with a second machine learning-based model executed by a receiver; and

sending, by the processor, the update packet to the receiver via a network, to update the second machine learning-based model.

2. The method as in claim 1 , further comprising:

receiving, at the processor, the telemetry data from the sub-system of the vehicle, wherein the sub-system of the vehicle comprises a Controller Area Network (CAN) bus.

3. The method as in claim 1 , wherein the second machine learning-based model is configured to predict physical characteristics of the vehicle for input to an application.

4. The method as in claim 1 , wherein detecting the difference between the physical characteristic predicted by the first model and the physical characteristic indicated by the telemetry data comprises:

determining whether a Bayesian error variance between the telemetry data and a state prediction by the first machine learning-based model exceeds a defined threshold.

5. The method as in claim 1 , wherein forming the packet payload of the update packet indicative of the detected difference, based in part on the relevancy of the physical characteristic to the first machine learning-based model, comprises:

determining the relevancy of the physical characteristic to the first machine learning-based model using a measure of covariance between the telemetry data and the first machine learning-based model.

6. The method as in claim 1 , wherein forming the packet payload of the update packet indicative of the detected difference, based in part on the relevancy of the physical characteristic to the first machine learning-based model, comprises:

adding data indicative of the detected difference to the packet payload; and

adding data to the packet payload that is indicative of one or more other detected differences between physical characteristics of the vehicle indicated by the telemetry data and physical characteristics of the vehicle predicted by the first machine learning-based model.

7. The method as in claim 1 , wherein applying the synchronization strategy to the update packet comprises:

including a sequence number in the update packet indicative of a sample count of the physical characteristic reported in the telemetry data, wherein the second machine learning-based model interpolates the physical characteristic a number of times according to the sequence number included in the update packet.

8. The method as in claim 1 , wherein applying the synchronization strategy to the update packet comprises:

including a sequence number in the update packet indicative of an ordering of update packets, wherein the second machine learning-based model interpolates the physical characteristic a number of times according to the sequence number included in the update packet.

9. The method as in claim 1 , wherein the processor and the receiver utilize synchronized clocks, and wherein applying the synchronization strategy to the update packet comprises:

including a timestamp in the update packet.

10. An apparatus, comprising:

one or more network interfaces to communicate with a network of a vehicle;

a processor coupled to the network interfaces and configured to execute one or more processes; and

a memory configured to store a process executable by the processor, the process when executed configured to:

detect a difference between a physical characteristic of the vehicle predicted by a first machine learning-based model and a physical characteristic of the vehicle indicated by telemetry data generated by a sub-system of the vehicle;

form a packet payload of an update packet indicative of the detected difference, based in part on a relevancy of the physical characteristic to the first machine learning-based model;

apply a synchronization strategy to the update packet, to synchronize the update packet with a second machine learning-based model executed by a receiver; and

send the update packet to the receiver via a network, to update the second machine learning-based model.

11. The apparatus as in claim 10 , wherein the process when executed is further configured to:

receive the telemetry data from the sub-system of the vehicle, wherein the sub-system of the vehicle comprises a Controller Area Network (CAN) bus.

12. The apparatus as in claim 10 , wherein the second machine learning-based model is configured to predict physical characteristics of the vehicle for input to an application.

13. The apparatus as in claim 10 , wherein the apparatus detects the difference between the physical characteristic predicted by the first model and the physical characteristic indicated by the telemetry data by:

determining whether a Bayesian error variance between the telemetry data and a state prediction by the first machine learning-based model exceeds a defined threshold.

14. The apparatus as in claim 10 , wherein the apparatus forms the packet payload of the update packet indicative of the detected difference, based in part on the relevancy of the physical characteristic to the first machine learning-based model, by:

determining the relevancy of the physical characteristic to the first machine learning-based model using a measure of covariance between the telemetry data and the first machine learning-based model.

15. The apparatus as in claim 10 , wherein the apparatus forms the packet payload of the update packet indicative of the detected difference, based in part on the relevancy of the physical characteristic to the first machine learning-based model, by:

adding data indicative of the detected difference to the packet payload; and

adding data to the packet payload that is indicative of one or more other detected differences between physical characteristics of the vehicle indicated by the telemetry data and physical characteristics of the vehicle predicted by the first machine learning-based model.

16. The apparatus as in claim 10 , wherein the apparatus applies the synchronization strategy to the update packet by:

including a sequence number in the update packet indicative of a sample count of the physical characteristic reported in the telemetry data, wherein the second machine learning-based model interpolates the physical characteristic a number of times according to the sequence number included in the update packet.

17. The apparatus as in claim 10 , wherein the apparatus applies the synchronization strategy to the update packet by:

including a sequence number in the update packet indicative of an ordering of update packets, wherein the second machine learning-based model interpolates the physical characteristic a number of times according to the sequence number included in the update packet.

18. The method as in claim 1 , wherein the processor and the receiver utilize synchronized clocks, and wherein the apparatus applies the synchronization strategy to the update packet by:

including a timestamp in the update packet.

19. The apparatus as in claim 10 , wherein the second machine learning-based model is executed by a cloud service that is remote from the vehicle.

20. A tangible, non-transitory, computer-readable medium storing program instructions that cause a processor in a vehicle to execute a process comprising:

detecting, by the processor of the vehicle, a difference between a physical characteristics of the vehicle predicted by a first machine learning-based model and a physical characteristic of the vehicle indicated by telemetry data generated by a sub-system of the vehicle;

forming, by the processor, a packet payload of an update packet indicative of the detected difference, based in part on a relevancy of the physical characteristic to the first machine learning-based model;

applying, by the processor, a synchronization strategy to the update packet, to synchronize the update packet with a second machine learning-based model executed by a receiver; and

sending, by the processor, the update packet to the receiver via a network, to update the second machine learning-based model.