Long range LIDAR-based speed estimation

#The present disclosure relates generally to the automotive, advanced driver assistance systems (ADAS), and autonomous driving (AD) fields. More particularly, the present disclosure relates to long range LIDAR-based speed estimation systems and methods.

#The absolute speed of an object in the vicinity of a vehicle is typically based on the determined self speed of the ego vehicle and the relative speed of the object, which may be another vehicle, for example. The self speed is typically determined from an external sensor, such as an inertial measurement unit (IMU), a global positioning system (GPS), or the vehicle odometer, which are known to persons of ordinary skill in the art. However, such sensors are greatly dependent upon other sensors and vehicle operating conditions and can be inaccurate. The relative speed is typically determined from camera images using annotation, object segmentation, and tracking, taking into account past trajectories. However, such object annotation and tracking is difficult when the object is at a long distance from the vehicle, as the object may be hard to detect and the camera can easily lose track of it. The result is error in the absolute speed measurement of surrounding objects, which can significantly impact vehicle safety in ADAS and AD environments.

#This, the present disclosure proposes the use of a LIDAR sensor to determine the absolute speed of an object in the vicinity of an ego vehicle. Advantageously, such LIDAR sensors have long range and are less affected by vehicle operating conditions.

#The present disclosure is illustrated and described with reference to the various drawings, in which like reference numbers are used to denote like system components/method steps, as appropriate, and in which:

#FIG. 1 is a schematic diagram illustrating one exemplary embodiment of the system and method for performing self speed estimation of the present disclosure;

#FIG. 2 is a schematic diagram illustrating one exemplary embodiment of the system and method for performing relative speed estimation of the present disclosure;

#The long range LIDAR-based speed estimation systems and methods of the present disclosure divide multiple LIDAR frames into ranges based on the distance of a point from the LIDAR sensor: short range (e.g. up to 50 or 70 m), middle range (e.g. from 50 or 70 m to 100 or 120 m), and long range (e.g. from 100 or 120 m to 250 m). Given these ranges, tasks are performed separately on each and the results are then combined. Advantageously, the systems and methods of the present disclosure are LIDAR agnostic and run in the 10 Hz region, for example. Two alternative methodologies are provided.

#In the first alternative methodology, both self-speed estimation and relative speed estimation are performed, with the combined result used to estimate absolute speed. Range affects are considered. Self speed estimation is performed for the short range; a relatively close range, the boundary of which is decided by benchmarking the self speed estimation. Relative speed estimation is performed for all ranges using a range awareness algorithm.

#In the second alternative methodology, a multi-task network shares both self speed estimation and relative speed estimation for all ranges. In terms of output, three regression heads produce two types of results: rotation and translation (for self speed estimation) and object location and speed (for relative speed estimation). The rotations, translations, and speeds produced by the multi-task network collectively provide the absolute speed of each object.