Predicting trajectories of objects based on contextual information
Aspects of the disclosure relate to detecting and responding to objects in a vehicle's environment. For example, an object may be identified in a vehicle's environment, the object having a heading and location. A set of possible actions for the object may be generated using map information describing the vehicle's environment and the heading and location of the object. A set of possible future trajectories of the object may be generated based on the set of possible actions. A likelihood value of each trajectory of the set of possible future trajectories may be determined based on contextual information including a status of the detected object. A final future trajectory is determined based on the determined likelihood value for each trajectory of the set of possible future trajectories. The vehicle is then maneuvered in order to avoid the final future trajectory and the object.
1. A computer-implemented method comprising:
receiving, by one or more computing devices, sensor data identifying a detected object in a vehicle's environment;
generating, by the one or more computing devices, a set of possible future trajectories of the detected object using map information;
determining, by the one or more computing devices, a likelihood value of each trajectory of the set of possible future trajectories based on contextual information for the detected object;
filtering the set of possible future trajectories based on the determined likelihood value of each trajectory of the set of possible trajectories;
determining, by the one or more computing devices, a final future trajectory based on the filtered set of possible future trajectories; and
maneuvering, by the one or more computing devices, the vehicle in order to avoid the final future trajectory and the detected object.
2. The method of claim 1 , wherein the filtering includes comparing the determined likelihood values of each trajectory of the set of possible trajectories to a threshold value.
3. The method of claim 2 , wherein determining the final future trajectory includes:
when none of the trajectories of the set of possible future trajectories meet the threshold value, identifying a plurality of waypoints for each trajectory in the set of trajectories, wherein a waypoint includes at least one of a position, a velocity, and a timestamp;
determining a trajectory of the vehicle, wherein the trajectory of the vehicle includes a plurality of waypoints; and
comparing, at a same timestamp, each of the waypoints to a waypoint associated with a trajectory of the vehicle in order to determine the final future trajectory.
4. The method of claim 1 , wherein the contextual information includes a type of the detected object.
5. The method of claim 4 , wherein the type of the detected object is a vehicle, a bicyclist, or a pedestrian.
6. The method of claim 1 , wherein the contextual information includes a status of a turn signal of the detected object.
7. The method of claim 1 , wherein the contextual information includes a status of a brake light of the detected object.
8. The method of claim 1 , wherein the contextual information includes at least one of a size or a shape of the detected object.
9. The method of claim 1 , wherein the contextual information includes a speed of the detected object.
10. The method of claim 1 , wherein the contextual information includes a heading of the detected object.
11. The method of claim 1 , wherein all trajectories of the filtered set of possible future trajectories are identified as final future trajectories, such that maneuvering the vehicle includes avoiding all of the trajectories of the filtered set of possible future trajectories.
12. The method of claim 1 , wherein determining the final future trajectory includes selecting a trajectory of the filtered set of possible future trajectories with a highest likelihood value as the final future trajectory.
13. The method of claim 1 , wherein the sensor data further identifies a second object in the vehicle's environment, and where determining the determined likelihood value of each trajectory of the set of possible future trajectories is further based on contextual information for the second object.
14. The method of claim 13 , wherein the contextual information for the second object includes a type of the second object.
15. The method of claim 13 , wherein the contextual information includes a status of a turn signal of the second object.
16. The method of claim 13 , wherein the contextual information includes a status of a brake light of the second object.
17. The method of claim 13 , wherein the contextual information includes at least one of a size or a shape of the second object.
18. The method of claim 13 , wherein the contextual information includes a speed of the second object.
19. The method of claim 13 , wherein the contextual information includes a heading of the second object.
20. A system comprising:
one or more computing devices configured to:
receive sensor data identifying a detected object in a vehicle's environment;
generate a set of possible future trajectories of the detected object using map information;
determine a likelihood value of each trajectory of the set of possible future trajectories based on contextual information for the detected object;
filter the set of possible future trajectories based on the determined likelihood value of each trajectory of the set of possible trajectories;
determine a final future trajectory based on the filtered set of possible future trajectories; and
maneuver the vehicle in order to avoid the final future trajectory and the detected object.