Modeling positional uncertainty of moving objects using precomputed polygons

1. A method comprising:

receiving, by a data processing system of an autonomous vehicle (AV), data related to an agent state of an object;

generating a polygon corresponding to the agent state;

identifying extreme vertices of the polygon along a longitudinal axis parallel to a heading direction of the object or along a lateral axis orthogonal to the heading direction;

applying, based on the extreme vertices, at least one expansion transformation to the polygon along the longitudinal axis or the lateral axis to generate a precomputed polygon; and

transmitting, to a control system of the AV to modify a driving path or speed of the AV, data representative of uncertainty in a longitudinal position or in a lateral position of the object computed based on the precomputed polygon.

2. The method of claim 1 , wherein identifying the extreme vertices of the polygon comprises:

identifying longitudinal extreme vertices of the polygon along the longitudinal axis; and

identifying lateral extreme vertices of the polygon along the lateral axis.

3. The method of claim 2 , wherein applying the at least one expansion transformation to the polygon along the longitudinal axis or the lateral axis comprises:

applying a first expansion transformation to the polygon along the longitudinal axis to duplicate each lateral extreme vertex; and

applying a second expansion transformation to the polygon along the lateral axis to duplicate each longitudinal extreme vertex.

4. The method of claim 1 , further comprising:

applying at least one additional expansion transformation to the precomputed polygon based on an uncertainty of the agent state, wherein a total number of vertices of the precomputed polygon is unchanged by the additional expansion transformation.

5. The method of claim 1 , wherein at least one additional expansion transformation is applied to the precomputed polygon responsive to the data processing system receiving a query from the control system of the AV, the query indicating a level of risk tolerance with respect to the agent state.

6. The method of claim 1 , further comprising:

computing a trajectory of the object comprising a series of agent states that includes the agent state of the object, wherein the precomputed polygon is used to model each agent state in the series of agent states.

7. A system comprising:

a memory; and

a data processing system of an autonomous vehicle (AV) communicatively coupled to the memory, wherein the data processing system is to:

receive data related to an agent state of an object;

generate a polygon corresponding to the agent state;

identify extreme vertices of the polygon along a longitudinal axis parallel to a heading direction of the object or along a lateral axis orthogonal to the heading direction;

apply, based on the extreme vertices, at least one expansion transformation to the polygon along the longitudinal axis or the lateral axis to generate a precomputed polygon; and

transmit, to a control system of the AV to modify a driving path or speed of the AV, data representative of uncertainty in a longitudinal position or in a lateral position of the object computed based on the precomputed polygon.

8. The system of claim 7 , wherein to identify the extreme vertices of the polygon, the data processing system is to:

identify longitudinal extreme vertices of the polygon along the longitudinal axis; and

identify lateral extreme vertices of the polygon along the lateral axis.

9. The system of claim 8 , wherein to apply the at least one expansion transformation to the polygon along the longitudinal axis or the lateral axis, the data processing system is to:

apply a first expansion transformation to the polygon along the longitudinal axis to duplicate each lateral extreme vertex; and

apply a second expansion transformation to the polygon along the lateral axis to duplicate each longitudinal extreme vertex.

10. The system of claim 7 , wherein the data processing system is to further:

apply at least one additional expansion transformation to the precomputed polygon based on an uncertainty of the agent state, wherein a total number of vertices of the precomputed polygon is unchanged by the additional expansion transformation.

11. The system of claim 7 , wherein at least one additional expansion transformation is to be applied to the precomputed polygon responsive to the data processing system receiving a query from the control system of the AV, the query indicating a level of risk tolerance with respect to the agent state.

12. The system of claim 7 , wherein the data processing system is to further:

compute a trajectory of the object comprising a series of agent states that includes the agent state of the object, wherein the precomputed polygon is to be used to model each agent state in the series of agent states.

13. A non-transitory computer-readable medium having instructions stored thereon that, when executed by a computing device, cause the computing device to:

receive data related to an agent state of an object;

generate a polygon corresponding to the agent state;

identify extreme vertices of the polygon along a longitudinal axis parallel to a heading direction of the object or along a lateral axis orthogonal to the heading direction;

apply, based on the extreme vertices, at least one expansion transformation to the polygon along the longitudinal axis or the lateral axis to generate a precomputed polygon; and

transmit, to a control system of an AV to modify a driving path or speed of the AV, data representative of uncertainty in a longitudinal position or in a lateral position of the object computed based on the precomputed polygon.

14. The non-transitory computer-readable medium of claim 13 , wherein to identify the extreme vertices of the polygon, the computing device is to:

identify longitudinal extreme vertices of the polygon along the longitudinal axis; and

identify lateral extreme vertices of the polygon along the lateral axis.

15. The non-transitory computer-readable medium of claim 14 , wherein to apply the at least one expansion transformation to the polygon along the longitudinal axis or the lateral axis, the computing device is to:

apply a first expansion transformation to the polygon along the longitudinal axis to duplicate each lateral extreme vertex; and

apply a second expansion transformation to the polygon along the lateral axis to duplicate each longitudinal extreme vertex.

16. The non-transitory computer-readable medium of claim 13 , wherein the computing device is to further:

apply at least one additional expansion transformation to the precomputed polygon based on an uncertainty of the agent state, wherein a total number of vertices of the precomputed polygon is unchanged by the additional expansion transformation.

17. The non-transitory computer-readable medium of claim 13 , wherein at least one additional expansion transformation is to be applied to the precomputed polygon responsive to the computing device receiving a query from the control system of the AV, the query indicating a level of risk tolerance with respect to the agent state.

18. The non-transitory computer-readable medium of claim 13 , wherein the computing device is to further:

compute a trajectory of the object comprising a series of agent states that includes the agent state of the object, wherein the precomputed polygon is to be used to model each agent state in the series of agent states.