CN-122001922-A - In-vehicle edge processing of time series data

Abstract

The present disclosure relates to on-board edge processing of time-series data. In an embodiment, a method includes receiving, by a vehicle control system of a vehicle, one or more first time series inputs associated with a first time window, wherein the one or more first time series inputs relate to one or more signals generated within the vehicle. The method also includes applying, by the vehicle control system, a first function to the one or more first time series inputs to generate a first value of a first metric for a first time window. The method also includes storing, by the vehicle control system, a first value of a first metric in the vehicle in association with a first time window in a first bin.

Inventors

• A. Prochte
• S. Baggas

Assignees

• 瑞维安知识产权控股有限责任公司

Dates

Publication Date

20260508

Application Date

20251110

Priority Date

20241108

Claims (20)

1. 1. A method for on-board edge processing of time series data, the method comprising: receiving, by a vehicle control system of a vehicle, one or more first time series inputs associated with a first time window, wherein the one or more first time series inputs relate to one or more signals generated within the vehicle; Applying, by the vehicle control system, a first function to the one or more first time series inputs to generate a first value of a first metric for the first time window, and The first value of the first metric in the vehicle is stored by the vehicle control system in a first bin in association with the first time window.
2. 2. The method of claim 1, wherein the first bin comprises a plurality of values of the first metric for a plurality of time windows of a period of time, the plurality of values comprising the first value of the first metric for the first time window.
3. 3. The method according to claim 2, wherein: The first bin implements a circular buffer configured to store a predetermined number of values of the first metric, and Storing the first value of the first metric includes overwriting at least one other value of the plurality of values of the first metric.
4. 4. The method of claim 2, wherein the plurality of time windows corresponds to a predetermined time resolution for the first metric.
5. 5. The method of claim 2, further comprising, by the vehicle control system: Receiving one or more second time series inputs associated with a second time window; Applying the first function to the one or more second time series inputs to generate a second value of the first metric for the second time window, and The second value of the first metric is stored in the first bin in association with the second time window.
6. 6. The method of claim 2, further comprising, by the vehicle control system: receiving a request for a value of the first metric over an aggregated window; aggregating the plurality of values of the first bin based on the aggregation window to generate a requested value, and The generated requested value is output.
7. 7. The method of claim 6, wherein the generated requested value is output to a machine learning model within the vehicle, the method further comprising running the machine learning model based on the generated requested value.
8. 8. The method of claim 1, further comprising, by the vehicle control system: applying a second function to the one or more first time series inputs to generate a value of a second metric for the first time window, and The value of the second metric is stored in a second bin in association with the first time window.
9. 9. The method of claim 8, wherein the second bin comprises a plurality of values of the second metric for a plurality of time windows of a period, the plurality of values comprising the value of the second metric for the first time window.
10. 10. The method of claim 1, further comprising discarding the one or more first time series inputs in response to the first value of the first metric being stored in the first bin.
11. 11. The method of claim 1, wherein the first value of the first metric comprises an aggregate value associated with the one or more first time series inputs.
12. 12. The method of claim 1, wherein the first value of the first metric comprises at least one of a maximum value, a minimum value, a sum, or a count of the one or more first time series inputs.
13. 13. A system for on-board edge processing of time series data, the system comprising: One or more of the memories may be provided, the one or more memories include executable instructions; and One or more processors in data communication with the one or more memories and configured to execute the executable instructions to: Receiving one or more first time series inputs associated with a first time window, wherein the one or more first time series inputs relate to one or more signals generated within a vehicle; applying a first function to the one or more first time series inputs to generate a first value of a first metric for the first time window, and The first value of the first metric in the vehicle is stored in a first bin in association with the first time window.
14. 14. The system of claim 13, wherein the first bin comprises a plurality of values of the first metric for a plurality of time windows of a period of time, the plurality of values comprising the first value of the first metric for the first time window.
15. 15. The system of claim 14, wherein: The first bin implements a circular buffer configured to store a predetermined number of values of the first metric, and The storing of the first value of the first metric includes overwriting at least one other value of the plurality of values of the first metric.
16. 16. The system of claim 14, wherein the one or more processors are further configured to execute the executable instructions to: Receiving one or more second time series inputs associated with a second time window; Applying the first function to the one or more second time series inputs to generate a second value of the first metric for the second time window, and The second value of the first metric is stored in the first bin in association with the second time window.
17. 17. The system of claim 14, wherein the one or more processors are further configured to execute the executable instructions to: receiving a request for a value of the first metric over an aggregated window; aggregating the plurality of values of the first bin based on the aggregation window to generate a requested value, and The generated requested value is output.
18. 18. The system of claim 17, wherein the generated requested value is output to a machine learning model within the vehicle, wherein the one or more processors are further configured to execute the executable instructions to run the machine learning model based on the generated requested value.
19. 19. The system of claim 13, wherein the one or more processors are further configured to execute the executable instructions to discard the one or more first time series inputs in response to the first value of the first metric being stored in the first bin.
20. 20. A computer program product comprising a non-transitory computer usable medium having a computer readable program code embodied therein, said computer readable program code adapted to be executed to implement a method comprising: Receiving one or more first time series inputs associated with a first time window, wherein the one or more first time series inputs relate to one or more signals generated within a vehicle; Applying, by the vehicle control system, a first function to the one or more first time series inputs to generate a first value of a first metric for the first time window, and The first value of the first metric in the vehicle is stored by the vehicle control system in a first bin in association with the first time window.

Description

In-vehicle edge processing of time series data Cross Reference to Related Applications The present application claims the benefit of priority from U.S. provisional patent application serial No. 63/718,443 filed on 8 at 11 at 2024, which is hereby incorporated by reference in its entirety as if fully and fully set forth herein. Background The present disclosure relates to edge processing. In particular, the present disclosure relates to techniques for processing time series data on an edge device such as a vehicle. Disclosure of Invention In an embodiment, one general aspect includes a method for on-board edge processing of time series data. The method includes receiving, by a vehicle control system of a vehicle, one or more first time series inputs associated with a first time window, wherein the one or more first time series inputs relate to one or more signals generated within the vehicle. The method also includes applying, by the vehicle control system, a first function to the one or more first time series inputs to generate a first value of a first metric for a first time window. The method also includes storing, by the vehicle control system, a first value of a first metric in the vehicle in association with a first time window in a first bin. In an embodiment, another general aspect includes a system for on-board edge processing of time series data. The system includes one or more memories having executable instructions. The system also includes one or more processors in data communication with the one or more memories and configured to execute the executable instructions to receive one or more first time series inputs associated with the first time window, wherein the one or more first time series inputs relate to one or more signals generated within the vehicle. The one or more processors are further configured to execute the executable instructions to apply a first function to the one or more first time series inputs to generate a first value of a first metric for a first time window. The one or more processors are further configured to execute the executable instructions to store a first value of a first metric in the vehicle in association with a first time window in a first bin. In an embodiment, another general aspect includes a computer program product comprising a non-transitory computer-usable medium having computer-readable program code embodied therein. The computer readable program code is adapted to be executed to implement a method. The method includes receiving one or more first time series inputs associated with a first time window, wherein the one or more first time series inputs relate to one or more signals generated within a vehicle. The method also includes applying a first function to one or more first time series inputs to generate a first value of a first metric for a first time window. The method also includes storing a first value of a first metric in the vehicle in a first bin in association with a first time window. Drawings FIG. 1A illustrates an example vehicle according to some embodiments. Fig. 1B illustrates a chassis of a vehicle according to certain embodiments. Fig. 2A is a schematic block diagram of components of a vehicle according to certain embodiments. Fig. 2B is a schematic block diagram of alternative components of a vehicle according to certain embodiments. FIG. 3 illustrates an example environment for a Telematics Control Module (TCM) in accordance with certain embodiments. Fig. 4 illustrates example operation of the TCM of fig. 3 according to some embodiments. FIG. 5 illustrates an example environment for estimating a vehicle battery charge time, according to some embodiments. FIG. 6 illustrates an example of a process for maintaining metric values based on time-series inputs within a vehicle, according to some embodiments. FIG. 7 illustrates an example of a process for responding to a query for metric values stored within a vehicle, in accordance with certain embodiments. Detailed Description Machine Learning (ML) models are increasingly implemented on edge devices to provide advanced functionality to such devices. For example, in a vehicle setting, the ML model can provide more accurate predictions of vehicle battery charge time relative to conventional physical models. However, implementing ML models on edge devices presents many challenges, as performing inference via ML models typically requires a significant amount of computational and bandwidth resources. One such challenge is the computational constraints of the edge device. To avoid such limitations, conventional approaches typically perform inference via cloud-based services. While such methods significantly reduce the computational and power requirements of the edge devices, they are less effective when performing inferences with large input data sets of time series data. For example, in a vehicle setting, a vehicle Electronic Control Unit (ECU) can generate a large amount of raw time-series da