Systems and methods for path planning of autonomous vehicles
This disclosure presents systems and methods for path planning in autonomous vehicles. The method involves a planner receiving perception data from a perception module, which includes tracking or predicted object data about obstacles in the environment of an autonomous vehicle. The tracking or predicted object data are determined based on high recall detection data and high precision detection data.
1 . A method of path planning by a planner of an autonomous vehicle, comprising: receiving by the planner perception data from a perception module, wherein the perception module is configured to:
generate high precision detection data based on data received from a set of sensors, wherein the received data represents objects or obstacles in an environment of the autonomous vehicle;
identify from the high precision detection data a first set of objects or obstacles that are classifiable by at least one known classifier;
track movement of one or more objects in the first set of objects or obstacles over time and maintain identity of the tracked one or more objects in the first set of objects or obstacles;
generate high recall detection data based on the received data;
identify from the high recall detection data a second set of objects or obstacles without using any classifier; and
filter out objects, from the second set of objects or obstacles, that correspond to the tracked one or more objects in the first set of objects or obstacles to obtain a filtered set of objects or obstacles,
wherein the perception data comprises tracking or predicted object data associated with objects or obstacles in the environment of the autonomous vehicle, and wherein the tracking or predicted object data are determined based on high recall detection data and high precision detection data;
generating by the planner a trajectory for controlling the autonomous vehicle based on the perception data received from the perception module; and
transmitting to a controller of the autonomous vehicle the trajectory such that the autonomous vehicle is navigated by the controller to a destination.
2 . The method of claim 1 , wherein the trajectory comprises instructions for the controller to maneuver the autonomous vehicle.
3 . The method of claim 2 , wherein the instructions comprise a throttle signal, a brake signal, a steering signal, or a combination thereof.
4 . The method of claim 1 , wherein the perception module is configured to:
perform an operation on the high precision detection data of the objects and the high recall detection data of the obstacles, based on a status of the autonomous vehicle or based on one or more characteristics of the objects or the obstacles.
5 . The method of claim 4 , wherein the operation comprises jointly optimizing the high precision detection data of the objects and the high recall detection data of the obstacles, when the autonomous vehicle is performing fallback maneuvers.
6 . The method of claim 4 , wherein the operation comprises determining a cover value between an obstacle in the high recall detection data and an object in the high precision detection data, by dividing area of intersection of the obstacle and the object by area of the obstacle.
7 . The method of claim 6 , wherein the operation comprises removing an obstacle from the high recall detection data if the cover value is greater than a threshold cover value.
8 . The method of claim 6 , wherein if a protrusion on the obstacle is only difference between the obstacle and the object, the operation comprises associating the protrusion with the object and removing the obstacle from the high recall detection data.
9 . The method of claim 6 , wherein if the cover value is smaller than the threshold cover value, the operation comprises maintaining respective representations of the obstacle and the object.
10 . The method of claim 6 , wherein if the cover value is smaller than the threshold cover value and if the obstacle is associated with the object, the operation comprises integrating one or more characteristics of the object into characteristics of the obstacle while maintaining respective representations of the obstacle and the object.
11 . The method of claim 1 , wherein the perception module is configured to:
determine attributes of each of the objects or obstacles based on the received data from the set of sensors;
integrate the attributes of each of the objects or obstacles; and
identify objects or obstacles based on the integrated attributes.
12 . The method of claim 1 , wherein the perception module is configured to generate the high precision detection data and the high recall detection data based on different subsets of data received from different sets of sensors.
13 . The method of claim 1 , wherein the perception module is configured to filter out a set of points corresponding to the tracked one or more objects in the first set of objects or obstacles from at least one point cloud of the second set of objects or obstacles.
14 . A system for controlling an autonomous vehicle, comprising a planner configured to:
receive perception data from a perception module, wherein the perception module is configured to:
generate high precision detection data based on data received from a set of sensors, wherein the received data represents objects or obstacles in an environment of the autonomous vehicle;
identify from the high precision detection data a first set of objects or obstacles that are classifiable by at least one known classifier;
track movement of one or more objects in the first set of objects or obstacles over time and maintain identity of the tracked one or more objects in the first set of objects or obstacles;
generate high recall detection data based on the received data;
identify from the high recall detection data a second set of objects or obstacles without using any classifier; and
filter out objects, from the second set of objects or obstacles, that correspond to the tracked one or more objects in the first set of objects or obstacles to obtain a filtered set of objects or obstacles,
wherein the perception data comprises tracking or predicted object data associated with objects or obstacles in the environment of the autonomous vehicle, and wherein the tracking or predicted object data are determined based on high recall detection data and high precision detection data;
generate a trajectory for controlling the autonomous vehicle based on the perception data received from the perception module; and
transmit to a controller of the autonomous vehicle the trajectory such that the autonomous vehicle is navigated by the controller to a destination.
15 . The system of claim 14 , wherein the trajectory comprises instructions for the controller to maneuver the autonomous vehicle.
16 . The system of claim 15 , wherein the instructions comprise a throttle signal, a brake signal, a steering signal, or a combination thereof.
17 . The system of claim 14 , wherein the perception module is configured to:
perform an operation on the high precision detection data of the objects and the high recall detection data of the obstacles, based on a status of the autonomous vehicle or based on one or more characteristics of the objects or the obstacles.
18 . The system of claim 17 , wherein the operation comprises jointly optimizing the high precision detection data of the objects and the high recall detection data of the obstacles, when the autonomous vehicle is performing fallback maneuvers.
19 . The system of claim 17 , wherein the operation comprises determining a cover value between an obstacle in the high recall detection data and an object in the high precision detection data, by dividing area of intersection of the obstacle and the object by area of the obstacle.
20 . The system of claim 19 , wherein the operation comprises removing an obstacle from the high recall detection data if the cover value is greater than a threshold cover value.
21 . The system of claim 19 , wherein if a protrusion on the obstacle is only difference between the obstacle and the object, the operation comprises associating the protrusion with the object and removing the obstacle from the high recall detection data.
22 . The system of claim 19 , wherein if the cover value is smaller than the threshold cover value, the operation comprises maintaining respective representations of the obstacle and the object.
23 . The system of claim 19 , wherein if the cover value is smaller than the threshold cover value and if the obstacle is associated with the object, the operation comprises integrating one or more characteristics of the object into characteristics of the obstacle while maintaining respective representations of the obstacle and the object.
24 . The system of claim 14 , wherein the perception module is configured to:
determine attributes of each of the objects or obstacles based on the received data from the set of sensors;
integrate the attributes of the each of the objects or obstacles; and
identify objects or obstacles based on the integrated attributes.
25 . The system of claim 14 , wherein the perception module is configured to generate the high precision detection data and the high recall detection data based on different subsets of data received from different sets of sensors.
26 . The system of claim 14 , wherein the perception module is configured to filter out a set of points corresponding to the tracked one or more objects in the first set of objects or obstacles from at least one point cloud of the second set of objects or obstacles.