Dynamic bounding box

Patent metadata

Field	Value
Publication number	US-12387345-B2
Application number	US-202217932056-A
Country	US
Kind code	B2
Filing date	Sep 14, 2022
Priority date	Sep 14, 2022
Publication date	Aug 12, 2025
Grant date	Aug 12, 2025

Abstract

A computer that includes a processor and a memory can determine a final trajectory for a vehicle by determining a candidate trajectory of a first object based on a detected second object. The candidate trajectory can be input to a reachable polyhedral marching processor to determine dynamic occupancy polyhedrals based on a shape of the candidate trajectory. A reachable tube can be determined based on combining the dynamic occupancy polyhedrals and the final trajectory can be determined based on the reachable tube avoiding the detected second object.

First claim

The invention claimed is:

1. A system, comprising:
a computer that includes a processor and a memory, the memory including instructions executable by the processor to determine a final trajectory by:
determining a candidate trajectory of a first object based on a detected second object;
inputting the candidate trajectory to a reachable polyhedral marching processor to determine dynamic occupancy polyhedrals based on a shape of the candidate trajectory wherein the sizes of the dynamic occupancy polyhedrals are based on estimates of errors;
determining a reachable tube based on combining the dynamic occupancy polyhedrals; and
determining the final trajectory based on the reachable tube avoiding the detected second object.

2. The system of claim 1 , wherein the instructions include further instructions to operate the first object based on the final trajectory by controlling actuators to control movement of the first object.

3. The system of claim 2 , wherein the first object is a vehicle and controlling the actuators to control the movement of the vehicle include controlling one or more of vehicle powertrain, vehicle steering and vehicle brakes.

4. The system of claim 2 , wherein the first object is a robot and controlling the actuators to control the movement of the robot include controlling movement of one or more of a gripper and a robotic arm.

5. The system of claim 1 , the instructions including further instructions to input the candidate trajectory to a first neural network to simplify the candidate trajectory.

6. The system of claim 5 , wherein the candidate trajectory is a T*n matrix that includes x and y locations, velocities, and heading angles at a plurality of time steps t and simplifying the trajectory reduces the T*n matrix to a vector with m dimensions, where m<T*n.

7. The system of claim 1 , wherein the reachable polyhedral marching processor is programmed based on weights output from a second neural network.

8. The system of claim 7 , wherein the second neural network determines density distributions based on solving Liouville partial differential equations.

9. The system of claim 8 , wherein the second neural network is trained based on a plurality of trajectories acquired from real world vehicles.

10. The system of claim 9 , wherein the second neural network includes fully connected neurons with ReLU activation and is trained based on stochastic gradient descent with L2-norm reconstruction loss.

11. The system of claim 1 , wherein the dynamic occupancy polyhedrals include two-dimensional regions wherein a probability that the first object will occupy locations within the reachable polyhedrals is higher than an empirically determined threshold.

12. The system of claim 1 , wherein the sizes of the dynamic occupancy polyhedrals are based on curvature and a distance between samples of segments of the candidate trajectory.

13. A method, comprising:
determining a final trajectory by:
determining a candidate trajectory of a first object based on a detected second object;
inputting the candidate trajectory to a reachable polyhedral marching processor to determine dynamic occupancy polyhedrals based on a shape of the candidate trajectory wherein the sizes of the dynamic occupancy polyhedrals are based on estimates of errors;
determining a reachable tube based on combining the dynamic occupancy polyhedrals; and
determining the final trajectory based on the reachable tube avoiding the detected second object.

14. The method of claim 13 , wherein the first object is operated based on the final trajectory by controlling one or more actuators to control movement of the first object.

15. The method of claim 14 , wherein the first object is a vehicle and controlling the actuators to control the movement of the vehicle includes controlling one or more of vehicle powertrain, vehicle steering and vehicle brakes.

16. The method of claim 14 , wherein the first object is a robot and controlling the actuators to control the movement of the robot includes controlling motion of one or more of a gripper and a robotic arm.

17. The method of claim 13 , further comprising inputting the candidate trajectory to a first neural network to simplify the candidate trajectory.

18. The method of claim 17 , wherein the candidate trajectory is a T*n matrix that includes x and y locations, velocities, and heading angles at a plurality of time steps t and simplifying the trajectory reduces the T*n matrix to a vector with m dimensions, where m<T*n.

19. The method of claim 13 , wherein the reachable polyhedral marching processor is programmed based on weights output from a second neural network.

20. The method of claim 19 , wherein the second neural network determines density distributions based on solving Liouville partial differential equations.

Assignees

• Ford Global Tech Llc
• Massachusetts Institute Of Tech Mit Tech Licensing Office
• Massachusetts Inst Technology

Inventors

• Waez Md Tawhid Bin
• Meng Yue
• Qiu Zeng
• Fan Chuchu

Classifications

• G06T2207/30241

  Trajectory · CPC title

• G06T2207/20084

  Artificial neural networks [ANN] · CPC title

• G06T2207/20081

  Training; Learning · CPC title

• B60W60/0027

  using trajectory prediction for other traffic participants · CPC title

• G06T2207/30252

  Vehicle exterior; Vicinity of vehicle · CPC title