Determining objects for position prediction in vehicles

1 . A system comprising:

one or more processors; and

one or more non-transitory computer-readable media storing instructions that, when executed, cause the system to perform operations comprising:

receiving sensor data from a sensor associated with a vehicle in an environment, the environment including a first object;

determining, based at least in part on the sensor data, a first set of positions reachable by the first object in a future time window, wherein the first set of positions are associated with a discretized array of positions associated with the environment;

determining, for a first position of the first set of positions, a time associated with arrival of the first object at the first position and a weight associated with the first position, the first position associated with a planned trajectory of the vehicle;

determining a time associated with arrival of the vehicle at the first position;

determining, based at least in part on the weight associated with the first position and the times associated with arrival of the first object and the vehicle at the first position, a score associated with the first position;

determining respective scores for a plurality of positions of the first set of positions;

determining, based at least in part on the respective scores of the plurality of positions of the first set of positions, a metric associated with the first object;

determining to include the first object in a set of objects based at least in part on the metric associated with the first object;

determining, based at least in part on a first prediction method, predicted future positions of objects in the set of objects, the first prediction method comprising one or more machine learned models; and

controlling the vehicle to traverse the environment based at least in part on the predicted future positions.

2 . The system of claim 1 , wherein the environment includes a second object, the operations comprising:

determining, based at least in part on the sensor data, a second set of positions reachable by the second object in the future time window;

determining a score associated with a second position of the second set of positions;

excluding the second object from the set of objects based at least in part on the score associated with the second position;

determining an estimated future position of the second object based at least in part on a second prediction method, the second prediction method associated with use of fewer computational resources than the first prediction method; and

controlling the vehicle based at least in on the estimated future position of the second object.

3 . The system of claim 1 , comprising determining at least one of: the first set of positions reachable by the first object, or the weight associated with the first position based at least in part on identifying the object as an on-road object or an off-road object.

4 . A method comprising:

receiving sensor data from a sensor associated with a vehicle in an environment, the environment including a first object;

determining, based at least in part on the sensor data, a set of reachable positions of the first object in a future time window;

determining, based at least in part on the set of reachable positions, a first metric associated with likelihood of the first object interacting with the vehicle in the future time window, wherein the set of reachable positions are associated with a discretized array of positions associated with the environment;

determining, based at least in part on the first metric, to predict a future position of the first object based at least in part on a first prediction method;

determining a trajectory for controlling the vehicle based at least in part on the future position of the first object; and

controlling the vehicle based at least in part on the trajectory.

5 . The method of claim 4 , comprising:

determining, based at least in part on the sensor data, a second metric associated with likelihood of a second object in the environment interacting with the vehicle in the future time window; and

determining, based at least in part on the second metric, to omit predicting a future position of the second object based at least in part on the first prediction method.

6 . The method of claim 5 , comprising:

determining, based at least in part on the second metric, an estimated future position of the second object based at least in part on a second prediction method.

7 . The method of claim 4 , comprising determining the first metric based at least in part on a classification associated with the first object.

8 . The method of claim 4 , wherein determining the set of reachable positions comprises:

determining a first subset of positions associated with a possible path of the first object; and

determining a second subset of positions comprising positions within a predetermined distance of the possible path of the first object.

9 . The method of claim 4 , comprising:

determining respective scores for plural positions of the set of reachable positions, a respective score associated with a likelihood of the vehicle interacting with the first object at the respective position;

wherein the first metric is determined based at least in part on the respective scores for the plural positions.

10 . The method of claim 9 , comprising determining the reachable set of positions of the first object based at least in part on a predetermined velocity or acceleration associated with one or more of: the first object, an object class associated with the first object, or the environment.

11 . The method of claim 9 , wherein a first score for a first position of the set of reachable positions is determined based at least in part on a time of arrival of the vehicle at the first position and a time of arrival of the first object at the first position.

12 . The method of claim 9 , wherein a second score for a second position of the set of reachable positions is determined based at least in part on one or more of:

(i) a number of lanes between the second position and a current position of the first object; or

(ii) an angle between the second position and a direction of a current velocity of the first object.

13 . The method of claim 9 , comprising omitting determining a score for a third position of the set of reachable positions based at least in part on one or more of:

(i) the third position not being associated with a planned trajectory of the vehicle;

(ii) a maximum arrival time of the vehicle at the third position being earlier than a minimum arrival time of the first object at the third position; or

(iii) a maximum arrival time of the first object at the second position being earlier than a minimum time of arrival of the vehicle at the third position.

14 . The method of claim 4 , comprising:

determining to predict the future position of the first object using the first prediction method based at least in part on the first metric satisfying a threshold criterion; or

determining to predict a future position of a third object using a second model based at least in part on a third metric associated with the third object not satisfying a threshold criterion.

15 . The method of claim 4 , comprising:

determining respective metrics associated with a plurality of objects in the environment;

determining to predict the future position of the first object using the first prediction method based at least in part on comparing the respective metrics of the plurality of objects; and

determining to predict the future position of a fourth object of the plurality of objects using a second prediction method based at least in part on comparing the respective metrics of the plurality of objects.

16 . One or more non-transitory computer-readable media storing instructions executable by one or more processors, wherein the instructions, when executed, cause the one or more processors to perform operations comprising:

receiving sensor data from a sensor associated with a vehicle in an environment, the environment including a first object;

determining, based at least in part on the sensor data, a set of reachable positions of the first object in a future time window;

determining, based at least in part on the set of reachable positions, a first metric associated with likelihood of the first object interacting with the vehicle in a future time window, wherein the set of reachable positions are associated with a discretized array of positions associated with the environment;

determining, based at least in part on the first metric, to predict a future position of the first object based at least in part on a first prediction method;

determining a trajectory for controlling the vehicle based at least in part on the future position of the first object; and

controlling the vehicle based at least in part on the trajectory.

17 . The one or more non-transitory computer-readable media of claim 16 , wherein the first object is a vehicle, a pedestrian, a cyclist, or an animal.