System and method for safe predictive control of autonomous vehicles
In variants, the system can include: a localization module; an objective generator; a mapping module; and a controller system including a set of Model Predictive Control (MPC) modules, a safety function generator, and a tire stiffness estimation module. The system can optionally include or be used with a Drive-by-Wire (DBW) system. The system functions to control an autonomous vehicle to track a planned trajectory. The system and/or components thereof are preferably located on a set of computing systems of an autonomous vehicle, but can additionally and/or alternatively be located elsewhere.
1 . A method, comprising:
capturing a set of measurements of a scene surrounding a vehicle;
based on the set of measurements, determining a trajectory for the vehicle;
based on the trajectory, determining a set of constraints for a pair of controllers, the pair of controllers comprising a lateral controller and a longitudinal controller, wherein the lateral controller and longitudinal controller determine vehicle commands in separate and distinct domains;
during a first timestep:
based on the trajectory and using a first controller of the pair of controllers, determining a first set of vehicle commands and a first set of coupling signals; and
based on the trajectory, the set of constraints, and the first set of coupling signals, and using a second controller of the pair of controllers, determining a second set of vehicle commands and a second set of coupling signals; and
during a second timestep:
based on the second set of coupling signals, and using the first controller, determining a third set of vehicle commands and a third set of coupling signals; and
based on the third set of coupling signals, using the second controller, determining a fourth set of vehicle commands;
determining a set of vehicle actuator control instructions based on the third set of vehicle commands and the fourth set of vehicle commands; and
controlling a set of actuators of the vehicle according to the set of vehicle actuator control instructions.
2 . The method of claim 1 , wherein the first controller is the longitudinal controller, wherein the first set of vehicle commands comprises a longitudinal kinematic parameter, and wherein the second controller is the lateral controller, wherein the second set of vehicle commands comprises an angular kinematic parameter.
3 . The method of claim 2 , wherein the first set of coupling signals and second set of coupling signals represent kinematic interdependencies between lateral and longitudinal vehicle dynamics.
4 . The method of claim 3 , wherein the first set of coupling signals comprises a longitudinal velocity of the vehicle, and the second set of coupling signals comprises a yaw rate and lateral velocity of the vehicle.
5 . The method of claim 1 , wherein the first and second set of vehicle commands comprise vehicle component controls.
6 . The method of claim 1 , wherein the set of constraints comprises a steering constraint having a 4 th -order control barrier function based on a lateral position error and an orientation error, and wherein determining the second set of vehicle commands comprises using the steering constraint.
7 . The method of claim 1 , wherein determining the set of constraints comprises:
based on a virtual representation of a set of objects in the scene, determining a set of regions within a trajectory error space defined by:
a first axis representing a translational error relative to the trajectory; and
a second axis representing a rotational error relative to the trajectory; and
transforming the set of regions into the set of constraints.
8 . The method of claim 7 , wherein the set of regions is a time series of multiple regions.
9 . The method of claim 7 , wherein the set of constraints comprises control barrier functions used by a strict subset of the pair of controllers during vehicle commands determination.
10 . The method of claim 7 , wherein determining the set of regions comprises generating a convex approximation of a plurality of error values in the trajectory error space.
11 . A method, comprising:
using a set of sensors onboard a vehicle, capturing a set of measurements of a set of objects proximal to the vehicle;
based on the set of measurements, determining a virtual representation of the set of objects;
based on the set of measurements, determining a vehicle trajectory;
based on the virtual representation of the set of objects, determining a set of regions within a trajectory error space defined by:
a first component of vehicle trajectory error; and
a second component of vehicle trajectory error distinct from the first component of vehicle trajectory error, wherein a region of the set of regions defines a constraint on error in the first component of vehicle trajectory as a function of the second component of vehicle trajectory error;
based on the set of regions in trajectory error space, determining a set of constraints;
based on the trajectory and the set of constraints, using a set of controllers, determining a set of vehicle controls; and
controlling a powertrain of the vehicle based on the set of vehicle controls.
12 . The method of claim 11 , wherein each region of the set of regions comprises a respective kinematic operating envelope for the vehicle relative to the trajectory.
13 . The method of claim 12 , wherein the first component of trajectory error is a lateral translational deviation from the trajectory, and the second component of trajectory error is a heading deviation from the trajectory.
14 . The method of claim 11 , wherein the set of regions comprises a time series of regions.
15 . The method of claim 14 , wherein determining the set of vehicle controls comprises using multiple distinct regions of the time series of regions.
16 . The method of claim 14 , wherein each region of the set of regions corresponds to a respective timestep in a control horizon.
17 . The method of claim 16 , wherein an object of the set of objects corresponds to the respective timestep, and detection of the respective object triggers determination of the respective region.
18 . The method of claim 14 , wherein a consecutive subset of regions of the time series of regions are identical.
19 . The method of claim 11 , wherein the set of controllers comprises:
a lateral controller determining commands for a first set of vehicle actuators; and
a longitudinal controller determining control instructions for a second set of vehicle actuators distinct from the first set of vehicle actuators.
20 . The method of claim 11 , wherein the set of constraints applies to exactly one controller of the set of controllers, wherein the exactly one controller is the lateral controller.