Predictive anti-lag braking control for autonomous driving
Devices, systems, and methods for predictive anti-lag braking control for autonomous vehicles are described. An example method of controlling a vehicle includes receiving, from a controller, a first plurality of values indicative of a brake wheel torque demand associated with a deceleration profile of the vehicle, deriving, based on the first plurality of values and a steady-state brake model, a second plurality of values indicative of a brake pressure demand, wherein an input to the steady-state brake model comprises the brake pressure demand and at least one previous brake wheel torque demand, and wherein an output of the steady-state brake system comprises a current brake wheel torque demand, generating, based on the second plurality of values and at least the steady-state model, a plurality of brake control commands, and controlling, using the plurality of brake control commands, the air brake system.
1. A method of controlling a vehicle with an air brake system, comprising:
receiving, from a controller, a first plurality of values indicative of a brake wheel torque demand associated with a deceleration profile of the vehicle;
generating a second plurality of values by adding an offset to the first plurality of values, wherein the offset corresponds to a response delay of the air brake system;
deriving, based on the second plurality of values and a steady-state brake model, a third plurality of values indicative of a brake pressure demand, wherein an input to the steady-state brake model comprises the brake pressure demand and at least one previous brake wheel torque demand, and wherein an output of the steady-state brake system comprises a current brake wheel torque demand;
generating, based on the third plurality of values and at least the steady-state model, a plurality of brake control commands; and
controlling, using the plurality of brake control commands, the air brake system,
wherein the generating the plurality of brake control commands is further based on a transient-state brake model in response to at least one of the third plurality of values being greater than a threshold,
wherein the generating the plurality of brake control commands is based on a blended brake model that is a weighted sum of the steady-state brake model and the transient-state brake model,
wherein the blended brake model is configured to minimize a tracking error of the brake wheel torque demand over a duration of the deceleration profile,
wherein minimizing the tracking error comprises iteratively solving a constrained boundary value problem using a quasi-Newton method or a secant method,
wherein the iteratively solving the constrained boundary value problem comprises using a barrier penalty that can adjusted using a scalar quantity.
2. The method of claim 1 , wherein the controller comprises a longitudinal dynamic model predictive control controller.
3. The method of claim 1 , wherein the generating the plurality of brake control commands excludes a use of a transient-state brake model in response to each of the third plurality of values being less than or equal to a threshold.
4. The method of claim 1 , wherein a weight of the weighted sum is based on the brake pressure demand.
5. The method of claim 1 , wherein the deceleration profile comprises a plurality of speed values, each of the speed values corresponding to a target speed at a future time instant.
6. A non-transitory computer readable medium storing executable instructions for a vehicle with an air brake system that, when executed by at least one processor, cause the at least one processor to perform at least the following:
receive a first plurality of values indicative of a brake wheel torque demand associated with deceleration profile of the vehicle;
derive, based on the first plurality of values and a steady-state brake model, a second plurality of values indicative of a brake pressure demand;
generate, based on the second plurality of values and a blended brake model, a plurality of brake control commands; and
control, using the plurality of brake control commands, the air brake system,
wherein the blended model is the steady-state model when each of the second plurality of values is less than a threshold and the blended model is a weighted sum of the steady-state brake model and a transient-state brake model when at least one of the second plurality of values is greater than the threshold,
wherein the transient-state brake model is a first-order lag model,
wherein an output of the transient-state brake model comprises a (k+1)-th lag-adjusted brake wheel torque demand, wherein an input to the transient-state model comprises a k-th lag-adjusted brake wheel torque demand, a k-th brake pressure demand, a derivative of the k-th brake pressure demand, and a k-th steady-state brake wheel torque demand, and wherein k is an integer indicative of a time step.
7. The non-transitory computer readable medium of claim 6 , wherein a weight of the weighted sum is based on a ratio of the k th brake pressure demand and a minimum brake pressure demand.
8. The non-transitory computer readable medium of claim 6 , wherein a rate of change of the lag-adjusted brake wheel torque demand is proportional to the brake pressure demand.
9. The non-transitory computer readable medium of claim 6 , wherein the steady-state brake model comprises a pressure-torque correlation derived using a linear least-squares method, and
wherein the steady-state model is configured to exclude one or more effects from road grade, driveline friction, or driveline disturbances in gear shifting, and wherein the steady-state model is configured to include a torque effect from a foundation air brake system.
10. The non-transitory computer readable medium of claim 6 , wherein the blended brake model is configured to minimize a tracking error of brake wheel torque delivery over the demand caused by a duration of a speed reduction profile.
11. An apparatus for a vehicle with an air brake system, comprising:
at least one processor; and
a memory including executable instructions that, when executed by the at least one processor, cause the apparatus to perform operations comprising:
receiving, from a controller, a first plurality of values indicative of a brake wheel torque demand associated with a deceleration profile of the vehicle;
generating a second plurality of values by adding an offset to the first plurality of values, wherein the offset corresponds to a response delay of the air brake system;
deriving, based on the second plurality of values and a steady-state brake model, a third plurality of values indicative of a brake pressure demand, wherein an input to the steady-state brake model comprises the brake pressure demand and at least one previous brake wheel torque demand, and wherein an output of the steady-state brake system comprises a current brake wheel torque demand;
generating, based on the third plurality of values and at least the steady-state model, a plurality of brake control commands; and
controlling, using the plurality of brake control commands, the air brake system,
wherein the generating the plurality of brake control commands is further based on a transient-state brake model in response to at least one of the third plurality of values being greater than a threshold,
wherein the generating the plurality of brake control commands is based on a blended brake model that is a weighted sum of the steady-state brake model and the transient-state brake model,
wherein the blended brake model is configured to minimize a tracking error of the brake wheel torque demand over a duration of the deceleration profile,
wherein minimizing the tracking error comprises iteratively solving a constrained boundary value problem using a quasi-Newton method or a secant method,
wherein the iteratively solving the constrained boundary value problem comprises:
checking, at an n-th iteration, whether an m-th variable representing an input to the n-th iteration exceeds a practical brake system limitation; and
truncating the m-th variable, upon determining that the m-th variable exceeds the practical brake system limitation, using a truncation rule,
wherein m and n are positive integers, and
wherein the truncation rule is based on an upper bound and a lower bound associated with the tracking error.
12. The apparatus for claim 11 , wherein the controller comprises a longitudinal dynamic model predictive control controller.
13. The apparatus for claim 11 , wherein the generating the plurality of brake control commands excludes a use of a transient-state brake model in response to each of the third plurality of values being less than or equal to a threshold.
14. The apparatus for claim 11 , wherein a weight of the weighted sum is based on the brake pressure demand.
15. The apparatus for claim 11 , wherein the deceleration profile comprises a plurality of speed values, each of the speed values corresponding to a target speed at a future time instant.