Autonomous vehicle actuation dynamics and latency identification
Systems and methods are disclosed for identifying time-latency and subsystem control actuation dynamic delay due to second order dynamics that are neglected in control systems of the prior art. Embodiments identify time-latency and subsystem control actuation delays by developing a discrete-time dynamic model having parameters and estimating the parameters using a least-squares method over selected crowd-driving data. After estimating the model parameters, the model can be used to identify dynamic actuation delay metrics such as time-latency, rise time, settling time, overshoot, bandwidth, and resonant peak of the control subsystem. Control subsystems can include steering, braking, and throttling.
1. A computer-implemented method of improving an autonomous driving vehicle (ADV) control system, comprising:
for each of one or more ADV control subsystems, generating a discrete-time dynamic model having a plurality of parameters, the discrete-time dynamic model characterizing an actuation delay of the ADV control subsystem;
estimating each of the plurality of parameters for the discrete-time dynamic model using crowd-sourced ADV driving data for a predetermined driving scenario;
determining a bandwidth metric of the ADV control subsystem using the discrete-time dynamic model with the estimated parameters; and
transforming the discrete-time dynamic model into a continuous-time dynamic model and determining a rise time metric and an overshoot metric using the continuous-time dynamic model, wherein the bandwidth, rise time, and overshoot metrics are utilized by an ADV subsystem controller to compensate the actuation delay for autonomously driving an ADV.
2. The method of claim 1 , further comprising determining a resonant peak metric of the ADV control subsystem using the discrete-time dynamic model, wherein the resonant peak metric is utilized by the ADV subsystem controller to compensate the actuation delay.
3. The method of claim 1 , further comprising determining a settling time metric for the ADV control subsystem using the continuous-time dynamic model, wherein the settling time metric is utilized by the ADV subsystem controller to compensate the actuation delay.
4. The method of claim 1 , wherein the actuation delay represents a difference between a time of an initial execution of a control command received by the ADV control subsystem and a time of a measured actuation of the ADV control subsystem in response to the initial execution of the control command.
5. The method of claim 1 , wherein the one or more ADV control subsystems comprise one or more of a throttle control subsystem, a braking control subsystem, or a steering control subsystem.
6. The method of claim 1 , wherein the crowd-sourced ADV driving data is selected according to one or more of: a specified ADV type, a specified ADV controller type, or one or more specified ADV subsystem controller types.
7. The method of claim 1 , wherein the discrete-time dynamic model for the ADV control subsystem is a second order dynamic model of the form:
subsystem
measured
(
z
)
=
b
1
z
+
b
2
z
2
+
a
1
z
+
a
2
subsystem
desired
(
z
)
,
wherein the plurality of parameters are a 1 , a 2 , b 1 , and b 2 .
8. The method of claim 1 , wherein generating the discrete-time dynamic model further comprises introducing a time-latency term for the ADV control subsystem, the time-latency term representing a time between transmitting of a command by an ADV controller and an initial execution of the command by the ADV subsystem controller.
9. The method of claim 1 , wherein estimating the plurality of parameters includes performing a least-squares fit of the crowd-sourced ADV driving data to determine the parameters of the discrete-time dynamic model for the ADV control subsystem.
10. A non-transitory machine-readable medium having instructions stored therein, which when executed by a processor, cause the processor to perform operations of improving an autonomous driving vehicle (ADV) control system, the operations comprising:
for each of one or more ADV control subsystems, generating a discrete-time dynamic model having a plurality of parameters, the discrete-time dynamic model characterizing an actuation delay of the ADV control subsystem;
estimating each of the plurality of parameters for the discrete-time dynamic model using crowd-sourced ADV driving data for a predetermined driving scenario;
determining a bandwidth metric of the ADV control subsystem using the discrete-time dynamic model with the estimated parameters; and
transforming the discrete-time dynamic model into a continuous-time dynamic model and determining a rise time metric and an overshoot metric using the continuous-time dynamic model, wherein the bandwidth, rise time, and overshoot metrics are utilized by an ADV subsystem controller to compensate the actuation delay for autonomously driving an ADV.
11. The machine-readable medium of claim 10 , wherein the operations further comprise determining a resonant peak metric of the ADV control subsystem using the discrete-time dynamic model, wherein the resonant peak metric is utilized by the ADV subsystem controller to compensate the actuation delay.
12. The machine-readable medium of claim 10 , wherein the operations further comprise determining a settling time metric for the ADV control subsystem using the continuous-time dynamic model, wherein the settling time metric is utilized by the ADV subsystem controller to compensate the actuation delay.
13. The machine-readable medium of claim 10 , wherein the actuation delay represents a difference between a time of an initial execution of a control command received by the ADV control subsystem and a time of a measured actuation of the ADV control subsystem in response to the initial execution of the control command.
14. The machine-readable medium of claim 10 , wherein the one or more ADV control subsystems comprise one or more of a throttle control subsystem, a braking control subsystem, or a steering control subsystem.
15. The machine-readable medium of claim 10 , wherein the crowd-sourced ADV driving data is selected according to one or more of: a specified ADV type, a specified ADV controller type, or one or more specified ADV subsystem controller types.
16. The machine-readable medium of claim 10 , wherein the discrete-time dynamic model for the ADV control subsystem is a second order dynamic model of the form:
subsystem
measured
(
z
)
=
b
1
z
+
b
2
z
2
+
a
1
z
+
a
2
subsystem
desired
(
z
)
,
wherein the plurality of parameters are a 1 , a 2 , b 1 , and b 2 .
17. The machine-readable medium of claim 10 , wherein generating the discrete-time dynamic model further comprises introducing a time-latency term for the ADV control subsystem, the time-latency term representing a time between transmitting of a command by an ADV controller and an initial execution of the command by the ADV subsystem controller.
18. The machine-readable medium of claim 10 , wherein estimating the plurality of parameters includes performing a least-squares fit of the crowd-sourced ADV driving data to determine the parameters of the discrete-time dynamic model for the ADV control subsystem.
19. 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 of improving an autonomous driving vehicle (ADV) control system, the operations including:
for each of one or more ADV control subsystems, generating a discrete-time dynamic model having a plurality of parameters, the discrete-time dynamic model characterizing an actuation delay of the ADV control subsystem;
estimating each of the plurality of parameters for the discrete-time dynamic model using crowd-sourced ADV driving data for a predetermined driving scenario;
determining a bandwidth metric of the ADV control subsystem using the discrete-time dynamic model with the estimated parameters; and
transforming the discrete-time dynamic model into a continuous-time dynamic model and determining a rise time metric and an overshoot metric using the continuous-time dynamic model, wherein the bandwidth, rise time, and overshoot metrics are utilized by an ADV subsystem controller to compensate the actuation delay for autonomously driving an ADV.
20. The system of claim 19 , wherein the operations further comprise determining a resonant peak metric of the ADV control subsystem using the discrete-time dynamic model, wherein the resonant peak metric is utilized by the ADV subsystem controller to compensate the actuation delay.