Model for excluding vehicle from sensor field of view
The technology relates to developing a highly accurate understanding of a vehicle's sensor fields of view in relation to the vehicle itself. A training phase is employed to gather sensor data in various situations and scenarios, and a modeling phase takes such information and identifies self-returns and other signals that should either be excluded from analysis during real-time driving or accounted for to avoid false positives. The result is a sensor field of view model for a particular vehicle, which can be extended to other similar makes and models of that vehicle. This approach enables a vehicle to determine when sensor data is of the vehicle or something else. As a result, the detailed modeling allowing the on-board computing system to make driving decisions and take other actions based on accurate sensor information.
1 . A method comprising:
receiving, by one or more processors of a vehicle, sensor data from one or more sensors of the vehicle for an external environment of the vehicle;
determining, by the one or more processors using a trained model having a range condition, which portions of the received sensor data should be filtered out;
generating, by the one or more processors, filtered sensor data based on the determining in accordance with the range condition; and
causing, by the one or more processors, the vehicle to operate in an autonomous driving mode based, at least in part, on the filtered sensor data.
2 . The method of claim 1 , wherein the model is trained based on an obstruction due to a deformable part of the vehicle.
3 . The method of claim 1 , wherein the model is trained based on an obstruction due to one or more transient items including dust, debris, gravel, water or snow spray.
4 . The method of claim 3 , wherein the model is trained to distinguish between the one or more transient items and pedestrians or other objects adjacent to the vehicle.
5 . The method of claim 1 , wherein the model is trained by evaluating arbitrary log data and determining which sensor data returns near the vehicle are associated with real objects that should be detected.
6 . The method of claim 1 , wherein the trained model is configured to perform sensor data self-return filtering.
7 . The method of claim 1 , wherein the sensor data includes a plurality of data points, the method further comprising evaluating the plurality of data points to determine whether or not the data points are sensor data self-return points or sensor data transient points.
8 . The method of claim 1 , wherein the trained model comprises a three-dimensional (3D) model of a given vehicle.
9 . The method of claim 8 , further comprising:
determining whether a range a sensor data point lies within a tolerance of a range predicted by the trained model to determine whether the sensor data point is a self-return data point; and
filtering out the sensor data point when the sensor data point is determined to be the self-return data point.
10 . The method of claim 9 , wherein when the range of the sensor data point is beyond a threshold value, the sensor data point is determined to be occluded by the vehicle.
11 . The method of claim 1 , wherein the sensor data is acquired while the vehicle is in a particular state.
12 . The method of claim 11 , wherein the particular state is based on whether a door of the vehicle is open or closed.
13 . The method of claim 11 , wherein the particular state is based on whether a trunk of the vehicle is open or closed.
14 . The method of claim 11 , wherein the particular state is based on an orientation of one or more wheels of the vehicle.
15 . The method of claim 11 , wherein the particular state is based on whether a hood of the vehicle is open or closed.
16 . A vehicle comprising:
one or more sensors; and
one or more processors configured to:
receive sensor data from the one or more sensors for an external environment of the vehicle;
determine, using a trained model having a range condition, which portions of the received sensor data should be filtered out;
generate filtered sensor data based on the determining in accordance with the range condition; and
cause the vehicle to operate in an autonomous driving mode based, at least in part, on the filtered sensor data.
17 . The vehicle of claim 16 , wherein the model is trained based on an obstruction due to a deformable part of the vehicle.
18 . The vehicle of claim 16 , wherein the sensor data includes a plurality of data points, and the one or more processors are further configured to evaluate the plurality of data points to determine whether or not the data points are sensor data self-return points or sensor data transient points.
19 . The vehicle of claim 16 , wherein the trained model comprises a three-dimensional (3D) model of a given vehicle.
20 . The vehicle of claim 19 , wherein the one or more processors are further configured to:
determine whether a range a sensor data point lies within a tolerance of a range predicted by the trained model to determine whether the sensor data point is a self-return data point; and
filter out the sensor data point when the sensor data point is determined to be the self-return data point.