Search

US-12617392-B1 - Simulation buildout with dynamic time intervals

US12617392B1US 12617392 B1US12617392 B1US 12617392B1US-12617392-B1

Abstract

Techniques for accurately predicting and avoiding collisions with objects detected in an environment of a vehicle are discussed herein. A vehicle computing device can determine a trajectory for a vehicle based on vehicle state data over a simulation period including a plurality of intervals. The vehicle computing device may determine the trajectory by determining control polic(ies) for at least a part of the simulation based on a control policy interval and determine control(s) and/or updated vehicle state data for at least the part of the simulation based on a current control policy and one or more dynamic integration interval.

Inventors

  • Weifu WANG

Assignees

  • Zoox, Inc.

Dates

Publication Date
20260505
Application Date
20231128

Claims (20)

  1. 1 . A method comprising: receiving vehicle state data of a vehicle; receiving an action for the vehicle to perform, the action associated with a set of controls; receiving a first data structure defining a set of control policy intervals and second data structure defining a set of integration intervals; determining, for a first control policy interval of the set of control policy intervals and based at least in part on the action, a modified control; and integrating, for a period of time based at least in part on the modified control and a first integration interval of the set of integration intervals, an updated vehicle state and an updated object state.
  2. 2 . The method of claim 1 , further comprising: determining the modified control based at least in part on predicting that controlling the vehicle in accordance with a first control of the set of controls is associated with at least a threshold probability of colliding with an object proximate the vehicle.
  3. 3 . The method of claim 2 , wherein the threshold probability of colliding with the object is based at least in part on recursively determining a predicted state of the object and a state of the vehicle.
  4. 4 . The method of claim 1 , wherein: the action comprises one or more of: continue along a lane, change lanes, or make a turn.
  5. 5 . The method of claim 1 , further comprising: distributing the first data structure defining the set of control policy intervals and the second data structure defining the set of integration intervals to a plurality of parallel processing units, wherein a first parallel processing unit determines the updated vehicle state and the updated object state associated with the action; and determining, by a second parallel processing unit of the plurality of parallel processing units, a second updated vehicle state and a second updated object state associated with a second action based on the first data structure defining the set of control policy intervals and the second data structure the set of integration intervals.
  6. 6 . The method of claim 1 , wherein at least one of: the first control policy interval differs from a second control policy interval of the set of control policy intervals, or the first integration interval of the set of integration intervals differs from a second integration interval of the set of integration intervals.
  7. 7 . One or more non-transitory computer-readable media storing instructions executable by one or more processors, wherein the instructions, when executed, cause the one or more processors to perform operations comprising: receiving vehicle state data of a vehicle; receiving an action for the vehicle to perform, the action comprising a set of controls; receiving a first data structure defining a set of control policy intervals and a second data structure a set of integration intervals; determining, for a first control policy interval of the set of control policy intervals and based at least in part on the action, a modified control; and integrating, for a period of time based at least in part on the modified control and a first integration interval of the set of integration intervals, one or more of an updated vehicle state or an updated object state.
  8. 8 . The one or more non-transitory computer-readable media of claim 7 , wherein the integrating comprises using at least one of: a physics policy, a dynamics policy, a kinematics policy, or a right of way policy, the right of way policy associated with one or more of: a roadway, an intersection, or a surface navigable by the vehicle.
  9. 9 . The one or more non-transitory computer-readable media of claim 7 , the operations further comprising: determining the modified control based at least in part on predicting that controlling the vehicle in accordance with a control of the set of controls is associated with at least a threshold probability of colliding with an object proximate the vehicle.
  10. 10 . The one or more non-transitory computer-readable media of claim 9 , wherein: wherein the threshold probability of colliding with the object is based at least in part on recursively determining a predicted state of the object and a state of the vehicle.
  11. 11 . The one or more non-transitory computer-readable media of claim 10 , wherein at least one of: the first control policy interval differs from a second control policy interval of the set of control policy intervals, or the first integration interval of the set of integration intervals differs from a second integration interval of the set of integration intervals.
  12. 12 . The one or more non-transitory computer-readable media of claim 11 , wherein one or more of the first control policy interval or an integration interval of the set of integration intervals is based at least in part on one or more of: a user input; a number of objects proximate the vehicle; an object classification of the object proximate the vehicle; an agent speed; geolocation; a power supply state of the vehicle; a temperature of a vehicle component; a speed of the vehicle; a proximity to an agent; or a risk assessment.
  13. 13 . The one or more non-transitory computer-readable media of claim 7 , wherein subsequent integration intervals of the set of integration intervals increase with respect to previous integration intervals.
  14. 14 . A system comprising: one or more processors; and one or more non-transitory computer-readable media storing instructions executable by the one or more processors, wherein the instructions, when executed, cause the system to perform operations comprising: receiving vehicle state data of a vehicle; receiving an action for the vehicle to perform, the action associated with a set of controls; receiving a first data structure defining a set of control policy intervals and second data structure defining a set of integration intervals; determining, for a first control policy interval of the set of control policy intervals and based at least in part on the action, a modified control; and integrating, for a period of time based at least in part on the modified control and a first integration interval of the set of integration intervals, an updated vehicle state and an updated object state.
  15. 15 . The system of claim 14 , the operations further comprising: determining the modified control based at least in part on predicting that controlling the vehicle in accordance with a first control of the set of controls is associated with at least a threshold probability of colliding with an object proximate the vehicle.
  16. 16 . The system of claim 15 , wherein the threshold probability of colliding with the object is based at least in part on recursively determining a predicted state of the object and a state of the vehicle.
  17. 17 . The system of claim 14 , wherein: the action comprises one or more of: continue along a lane, change lanes, or make a turn.
  18. 18 . The system of claim 14 , the operations further comprising: distributing the first data structure defining the set of control policy intervals and the second data structure defining the set of integration intervals to a plurality of parallel processing units, wherein a first parallel processing unit determines the updated vehicle state and the updated object state associated with the action; and determining, by a second parallel processing unit of the plurality of parallel processing units, a second updated vehicle state and a second updated object state associated with a second action based on the first data structure defining the set of control policy intervals and the second data structure the set of integration intervals.
  19. 19 . The system of claim 14 , wherein at least one of: the first control policy interval differs from a second control policy interval of the set of control policy intervals, or a first integration interval of the set of integration intervals differs from a second integration interval of the set of integration intervals.
  20. 20 . The system of claim 14 , wherein the integrating comprises using at least one of: a physics policy, a dynamics policy, a kinematics policy, or a right of way policy, the right of way policy associated with one or more of: a roadway, an intersection, or a surface navigable by the vehicle.

Description

BACKGROUND Planning systems in autonomous and semi-autonomous vehicles may determine actions for a vehicle to take in an operating environment. Actions for a vehicle may be determined based in part on avoiding objects present in the environment. For example, an action may be generated by a planning system to yield to a pedestrian, to change a lane to avoid another vehicle in the road, or the like. Perception systems utilize sensor data from sensors to “see” the environment, which enables the planning systems to determine an effect of a detected object on a potential action for the vehicle. BRIEF DESCRIPTION OF THE DRAWINGS The detailed description is described with reference to the accompanying figures. In the figures, the left-most digit(s) of a reference number identifies the figure in which the reference number first appears. The use of the same reference numbers in different figures indicates similar or identical components or features. FIG. 1 is an illustration of an example environment, in which an example vehicle applies a model to predict an intersection value indicating a likelihood for collision with one or more example objects. FIG. 2 is an illustration of another example environment in which one or more models determine potential vehicle states and/or object states at a future time. FIG. 3A is an illustration of an example dynamic time interval timing structure which may indicate the time intervals for a control policy component and an integration component. FIG. 3B is an illustration of an example data structures that may define a dynamic interval timing of a simulation buildout for a buildout component. FIG. 4 is a block diagram of an example system for implementing the techniques described herein. FIG. 5 is an illustration of a flowchart depicting an example process for determining a predicted trajectory using dynamic time intervals. FIG. 6 is a flowchart depicting an example process for controlling a vehicle based on a candidate trajectory determined using dynamic timing intervals. DETAILED DESCRIPTION Techniques for predicting trajectories of objects and/or a vehicle using dynamic time interval(s) or dynamic numbers of tick(s) (e.g., discrete step(s) in time) are discussed herein. The techniques can include a vehicle computing device that implements a model to determine realistic interactions between multiple objects and improve processing when determining an action for an autonomous vehicle. For instance, the model may receive a set of predicted object trajectories associated with an object along with one or more potential actions for the autonomous vehicle (e.g., stay in current lane, change lane to the right, change lane to the left, turn right at an intersection, pass one or more vehicles, etc.) and determine one or more predicted trajector(ies) for the potential actions based on the current state of the environment. More particularly, the model may, for an action of the autonomous vehicle, perform a simulation buildout operation to buildout a simulation for the action and/or an action which transitions from tracking a first action to a second action over a number of ticks (e.g., a simulation over 40 0.1 seconds ticks (four seconds total)). For each tick, the model may perform one, both or neither of a control policy function of a control policy component to determine control(s) for the autonomous vehicle (e.g., determine one or more of controls for the action (e.g., a steering control, a braking control, an acceleration control, and so on)) and/or perform an integration operation of an integration component to integrate the state of the environment based on the current controls output by the control policy function (e.g., to determine a next future state of the environment based on a current state and the controls determined by the control policy component). Further, a buildout component of a vehicle may use dynamic time interval(s) to determine which time steps or ticks within a simulation buildout the various operations may be performed (e.g., operations of a control policy component and an integration component). The various operations, such as operations of a control policy component and an integration component, may be performed multiple times over the course of the simulation buildout (e.g., at various ticks or points in time within the simulation buildout), resulting in a predicted trajectory or trajectories for the autonomous vehicle and other objects in the environment for the current state of the environment and the action. For example, the control policy component may operate to determine control(s) for the vehicle every four (4) ticks (e.g., at 0.4 seconds time intervals) over the eight second simulation buildout. Similarly, the integration component may determine the predicted responses of the vehicle at every two ticks (0.2 seconds) over the eight second simulation buildout. More particularly, the integration component may determine, for a current tick in the simulation b