Cross-platform control profiling tool for autonomous vehicle control

Patent metadata

Field	Value
Publication number	US-11511760-B2
Application number	US-202016643154-A
Country	US
Kind code	B2
Filing date	Jan 23, 2020
Priority date	Jan 23, 2020
Publication date	Nov 29, 2022
Grant date	Nov 29, 2022

Abstract

Systems and methods are disclosed for collecting driving data from simulated autonomous driving vehicle (ADV) driving sessions and real-world ADV driving sessions. The driving data is processed to exclude manual (human) driving data and to exclude data corresponding to the ADV being stationary (not driving). Data can further be filtered based on driving direction: forward or reverse driving. Driving data records are time stamped. The driving data can be aligned according to the timestamp, then a standardized set of metrics is generated from the collected, filtered, and time-aligned data. The standardized set of metrics are used to grade the performance the control system of the ADV, and to generate an updated ADV controller, based on the standardized set of metrics.

First claim

What is claimed is:

1. A computer-implemented method of improving an autonomous driving vehicle (ADV) control system, comprising:
extracting, from driving records of a plurality of ADVs including a simulated ADV and a real ADV, driving records for a specified ADV type having a specified ADV controller type which includes a model predictive control (MPC) controller type, a linear quadratic regulator (LQR) controller type, or a hybrid controller type;
filtering the extracted driving records to exclude driving records representing manual driving and to exclude driving records representing a stationary state of an ADV in the plurality of ADVs;
generating a set of standardized metrics, each metric representing a performance characteristic of an ADV controller having the specified ADV controller type for the specified ADV type;
using the set of standardized metrics to generate an updated ADV controller having the specified ADV controller type;
performing a simulation with the simulated ADV with the updated ADV controller to test the updated ADV controller;
distributing the updated ADV controller of the specified ADV controller type based on whether the specified ADV controller type is the MPC controller type, the LQR controller type, or the hybrid controller type, to one or more real ADVs of the ADV type, for use in driving the one or more ADVs; and
distributing the updated ADV controller to the simulated ADV to perform one or more additional simulations with.

2. The method of claim 1 , wherein the simulated ADV has an ADV type and an ADV controller type corresponding to the real ADV in the plurality of ADVs.

3. The method of claim 1 , wherein the extracted driving records are further filtered to include driving records corresponding to a gearshift position of the ADV, wherein the gearshift position comprises one of a forward gear or a reverse gear position.

4. The method of claim 1 , wherein the stationary state of the ADV is determined by at least one of:
a gearshift position of the ADV being one of neutral or park; or
a speed of the ADV being zero.

5. The method of claim 1 , further comprising:
aligning the extracted driving records for each ADV in the plurality of ADVs according to a timestamp of each extracted driving record.

6. The method of claim 1 , wherein, for each ADV in the plurality of ADVs, the extracted records of driving data comprise:
controller outputs to controlled systems of the ADV, including throttle, brakes, and steering, and including measured states of the controlled systems responsive to the controller outputs;
state information of the chassis of the ADV, including pitch, roll, ADV incline, and forward and lateral acceleration; and
location information indicating a planned location and heading of the ADV and an actual location and heading of the ADV, at a timestamp of the location information.

7. The method of claim 1 , further comprising:
generating visualization information from the extracted data records; and
providing the visualization information to at least one of:
a system that generates a report including plots and histograms of the visualization information; or
a simulation system that generates one or more charts, histograms, or plots for display on the simulation system, using the visualization information.

8. The method of claim 1 , further comprising:
generating one or more suggested updates to the ADV controller, based upon the generated set of standardized metrics.

9. The method of claim 8 , wherein the simulated ADV is run on a simulation system which generates a second set of standardized metrics based upon a second set of driving records generated by performing the simulation or the one or more additional simulations.

10. A non-transitory machine-readable medium having instructions stored therein, which when executed by a processor, cause the processor to perform operations, the operations comprising:
extracting, from driving records of a plurality of ADVs including a simulated ADV and a real ADV, driving records for a specified ADV type having a specified ADV controller type which includes a model predictive control (MPC) controller type, a linear quadratic regulator (LQR) controller type, or a hybrid controller type;
filtering the extracted driving records to exclude driving records representing manual driving and to exclude driving records representing a stationary state of an ADV in the plurality of ADVs;
generating a set of standardized metrics, each metric representing a performance characteristic of an ADV controller having the specified ADV controller type for the specified ADV type;
using the set of standardized metrics to generate an updated ADV controller having the specified ADV controller type;
performing a simulation with the simulated ADV with the updated ADV controller to test the updated ADV controller;
distributing the updated ADV controller of the specified ADV controller type based on whether the specified ADV controller type is the MPC controller type, the LQR controller type, or the hybrid controller type, to one or more ADVs of the ADV type, for use in driving the one or more ADVs; and
distributing the updated ADV controller to the simulated ADV to perform one or more additional simulations with.

11. The medium of claim 10 , wherein the simulated ADV has ADV type and a specified controller type corresponding to the real ADV in the plurality of ADVs.

12. The medium of claim 10 , wherein the extracted driving records are further filtered to include driving records corresponding to a gearshift position of the ADV, wherein the gearshift position comprises one of a forward gear or a reverse gear position.

13. The medium of claim 10 , the operations further comprising:
aligning the extracted driving records for each ADV in the plurality of ADVs according to a timestamp of each extracted driving record, wherein, for each ADV in the plurality of ADVs, the extracted records of driving data comprise:
controller outputs to controlled systems of the ADV, including throttle, brakes, and steering, and including measured states of the controlled systems responsive to the controller outputs;
state information of the chassis of the ADV, including pitch, roll, ADV incline, and forward and lateral acceleration; and
location information indicating a planned location and heading of the ADV and an actual location and heading of the ADV, at a timestamp of the location information.

14. The medium of claim 10 , the operations further comprising:
generating visualization information from the extracted data records; and
providing the visualization information to at least one of:
a system that generates a report including plots and histograms of the visualization information; or
a simulation system that generates one or more charts, histograms, or plots for display on the simulation system, using the visualization information.

15. The medium of claim 10 , the operations further comprising:
generating one or more suggested updates to the ADV controller, based upon the generated set of standardized metrics wherein
the simulated ADV runs on a simulation system using the updated ADV controller;
and generates a second set of standardized metrics based upon a second set of driving records generated by performing the simulation or the one or more additional simulations.

16. A data processing system, comprising:
a processor; and
a memory coupled to the processor to store instructions, which when executed by the processor, cause the processor to perform operations, the operations including:
extracting, from driving r

Assignees

• Baidu Usa Llc
• Baidu Com Times Tech Beijing Co Ltd

Inventors

• Wang Yu
• Luo Qi
• Cao Yu
• Feng Zongbao
• Lin Longtao
• Xiao Xiangquan
• Miao Jinghao
• Hu Jiangtao
• Wang Jingao
• Jiang Shu
• Zhou Jinyun
• Xu Jiaxuan

Classifications

• G06F16/2322

  using timestamps · CPC title

• B60W2050/0075

  Automatic parameter input, automatic initialising or calibrating means · CPC title

• B60W2556/10

  Historical data · CPC title

• B60W2520/125

  Lateral acceleration · CPC title

• B60W40/105 Primary

  Speed · CPC title