Granted Patent B2

US 12,505,675 · App. 17/852,015 · Granted Dec 23, 2025

Inventors: Juan Cheng (Sunnyvale, CA); Thomas Sterling (Lakewood, CO); Ulrich Vollath (Ismaning, DE); Brandon Sights (San Marcos, CA)

Assignee: Trimble Inc.

G06V20/58G01S13/865G01S13/89G01S17/89

Abstract

Claims (39)

1 . A method of identifying obstacles during navigation of a moving machine, the method comprising: capturing first radar sensor data from a radar sensor;

filtering the first radar sensor data using radar filter parameters to produce first filtered radar sensor data;

producing first radar obstacle points from the first filtered radar sensor data;

capturing lidar sensor data from a lidar sensor;

producing lidar obstacle points from the lidar sensor data; and optimizing the radar filter parameters for a signal-to-noise (SNR) filter and a spatial-temporal filter using the lidar obstacle points to produce optimized radar filter parameters, wherein the optimizing is performed by sweeping through SNR parameters first, then spatial filter parameters, then temporal parameters and comprises:

determining spatial filter bounding areas for obstacles identified by the lidar obstacle points, identifying the first radar obstacle points inside the spatial filter bounding areas as true positives, and tuning the radar filter parameters to maximize a number of true positives and produce the optimized radar filter parameters, capturing second radar sensor data from the radar sensor;

filtering the second radar sensor data using the optimized radar filter parameters to produce second filtered radar sensor data;

producing second radar obstacle points from the second filtered radar sensor data; and

using the second radar obstacle points during navigation of the moving machine.

2 . The method of claim 1 wherein the temporal parameters include a minimum delta time, a maximum time decay, a time decay rate, a first ring scale and a second ring scale.

3 . The method of claim 1 wherein additional filter parameters include a maximum distance, a grid width, and a grid depth.

4 . The method of claim 1 wherein obstacle points outside an area of interest and not within both a radar FOV and a lidar FOV are not processed.

5 . The method of claim 1 wherein the optimizing includes determining a F1 score as a harmonic mean of precision and recall;

maximizing both precision and recall; and

applying a beta factor that shifts an importance in favor of recall.

6 . The method of claim 1 wherein the radar filter parameters are optimized using one of Bayesian Optimization, Evolutionary Optimization, Simulated Annealing and Quadratic Optimization.

7 . An apparatus comprising:

a radar sensor mounted on a moving machine;

a lidar sensor mounted on the moving machine;

a processor mounted on the moving machine;

a memory mounted on the moving machine and containing non-transitory, computer readable media with instructions that, when executed by a computer, perform steps comprising:

capturing radar sensor data from a radar sensor;

filtering the radar sensor data using radar filter parameters to produce filtered radar sensor data;

producing radar obstacle points from the filtered radar sensor data;

capturing lidar sensor data from a lidar sensor;

producing lidar obstacle points from the lidar sensor data; and

optimizing the radar filter parameters using the lidar obstacle points to produce optimized radar filter parameters, wherein the optimizing comprises:

determining spatial filter bounding areas for obstacles identified by the lidar obstacle points;

identifying the radar obstacle points inside the spatial filter bounding areas as true positives;

tuning the radar filter parameters to maximize a number of true positives and produce the optimized radar filter parameters;

determining a F1 score as a harmonic mean of precision and recall;

maximizing both precision and recall; and

applying a beta factor that shifts an importance in favor or recall;

wherein the apparatus is configured to use the optimized radar filter parameters during navigation of the moving machine.

8 . The apparatus of claim 7 wherein the radar sensor filter parameters are optimized for a signal-to-noise (SNR) filter and a spatial-temporal filter.

9 . (The apparatus of claim 8 wherein the optimization is performed by sweeping through SNR parameters first, then spatial filter parameters, then temporal parameters.

10 . The apparatus of claim 9 wherein the temporal parameters include a minimum delta time, a maximum time decay, a time decay rate, a first ring scale and a second ring scale.

11 . The apparatus of claim 9 wherein additional filter parameters include a maximum distance, a grid width, and a grid depth.

12 . The apparatus of claim 7 wherein obstacle points outside an area of interest and not within both a radar FOV and a lidar FOV are not processed.

Assignments (2)

Continuity (1)

References Cited (39)