Prediction model with variable time steps
AssigneeZoox, Inc.
InventorEthan Miller Pronovost
Life of this patent
Estimating the term…
Abstract
Techniques for implementing a machine learned model to determine variable step sizes for searching a decision tree are described herein. For example, the techniques may include a computing device implementing a tree search algorithm to search the decision tree comprising nodes representing different possible object actions and/or vehicle actions at a future time. A search of the decision tree can be performed using multiple steps associated with different time periods to output a vehicle trajectory for the vehicle. The vehicle trajectory can be sent to a vehicle computing device for consideration during vehicle planning, which may include simulation.
Claims20
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 one or more processors to perform operations comprising:
- receiving, by a prediction component of an autonomous vehicle in an environment, from a planning component of the autonomous vehicle, a request for discrete predictions associated with the autonomous vehicle within a future time horizon;
- determining, by the prediction component, a first discrete prediction associated with the autonomous vehicle based at least in part on searching a decision tree wherein the first discrete prediction corresponds to a first time step change of the environment within the future time horizon;
- determining, by the prediction component and based at least in part on the first discrete prediction, a second discrete prediction associated with the autonomous vehicle wherein the second discrete prediction corresponds to a second time step change of the environment within the future time horizon, the second time step different form the first time step;
- and
- transmitting data representing the first discrete prediction and the second discrete prediction to the planning component to control the autonomous vehicle in the environment.
One or more non-transitory computer-readable media storing instructions that, when executed, cause one or more processors to perform operations comprising:
- receiving, by a model, a request for a prediction associated with a vehicle and an object in an environment;
- determining, by the model, first state data associated with the vehicle or the object based at least in part on searching a decision tree, wherein the first state data corresponding to a first time period;
- determining, by the model and based at least in part on the first state data, second state data associated with the vehicle or the object over a second time period, wherein the first time period is different from the second time period;
- and
- transmitting data representing the second state data to a component of a vehicle computing device to control the vehicle in the environment.
A method comprising:
- receiving, by a model, a request for a prediction associated with a vehicle and an object in an environment;
- determining, by the model, first state data associated with the vehicle or the object based at least in part on searching a decision tree wherein the first state data corresponds to a first time period;
- determining, by the model and based at least in part on the first state data, second state data associated with the vehicle or the object over a second time period, wherein the first time period is different from the second time period;
- and
- transmitting data representing the second state data to a component of a vehicle computing device to control the vehicle in the environment.
Dependencies are inferred from the claim text.