Image space motion planning of an autonomous vehicle
An autonomous vehicle that is equipped with image capture devices can use information gathered from the image capture devices to plan a future three-dimensional (3D) trajectory through a physical environment. To this end, a technique is described for image-space based motion planning. In an embodiment, a planned 3D trajectory is projected into an image-space of an image captured by the autonomous vehicle. The planned 3D trajectory is then optimized according to a cost function derived from information (e.g., depth estimates) in the captured image. The cost function associates higher cost values with identified regions of the captured image that are associated with areas of the physical environment into which travel is risky or otherwise undesirable. The autonomous vehicle is thereby encouraged to avoid these areas while satisfying other motion planning objectives.
1. A system comprising:
a processing unit; and
a memory unit coupled to the processing unit, the memory unit including instructions stored thereon, which when executed by the processing unit, cause the system to:
generate a cost function map that associates a cost value with each of multiple regions of an image of a physical environment,
wherein the cost value is indicative of a level of risk associated with navigating in an area of the physical environment corresponding to the region of the image;
identify, based on the cost function map, a particular region of the multiple regions of the image that is associated with a higher cost value than one or more other regions of the multiple regions of the image; and
generate a 3D trajectory for an autonomous vehicle through the physical environment that avoids the particular region of the image to lesson an overall cost of the 3D trajectory.
2. The system of claim 1 , further comprising:
a propulsion system; and
a flight controller configured to:
process the 3D trajectory to generate control commands; and
execute the control commands to cause the propulsion system to maneuver the autonomous vehicle along the 3D trajectory.
3. The system of claim 1 , further comprising:
a camera configured to capture the image of the physical environment.
4. The system of claim 1 , wherein the autonomous vehicle is an unmanned aerial vehicle (UAV).
5. The system of claim 1 , wherein to generate the cost function map, the instructions, when executed by the processing unit, cause the system to:
feed the image into a machine learning model to determine the level of risk associated with navigating in the area of the physical environment corresponding to the region of the image.
6. The system of claim 1 , wherein the instructions, when executed by the processing unit, cause the system to:
determine a cost value associated with a particular region each of the multiple regions of the image.
7. The system of claim 6 , wherein to determine the cost value associated with a particular region of the multiple regions of the image, the instructions, when executed by the processing unit, cause the system to:
estimate depth to a physical surface in an area of the physical environment corresponding to the particular region of the image; and
determine a level of confidence in the estimated depth.
8. The system of claim 1 , wherein the instructions, when executed by the processing unit, cause the system to:
receive an additional image of the physical environment while the vehicle is in motion through the physical environment;
process the additional image to dynamically update the cost function map.
9. The system of claim 8 , wherein the instructions, when executed by the processing unit, cause the system to:
identify, based on the updated cost function map, a particular region of the additional image that is associated with a higher cost value than one or more other regions of the additional image; and
optimize the 3D trajectory for an autonomous vehicle through the physical environment to lessen the overall cost of the 3D trajectory by avoiding the particular region of the additional image.
10. A method comprising:
generating a cost function map that associates a cost value with each of multiple regions of an image of a physical environment, wherein each cost value is indicative of a level of risk associated with navigating in an area of the physical environment corresponding to the associated region; and
optimizing, using the cost function map, a 3D trajectory for an autonomous vehicle through the physical environment by reducing an overall cost of the 3D trajectory.
11. The method of claim 10 , further comprising:
identifying, based on the cost function map, a particular region of the image that is associated with a higher cost value than one or more other regions of the multiple regions of the image;
wherein the 3D trajectory for the autonomous vehicle is optimized by adjusting the 3D trajectory such that a projection of the 3D trajectory in an image space of the image avoids the particular region of the image.
12. The method of claim 10 , further comprising:
processing the 3D trajectory to generate control commands; and
controlling a propulsion system of the autonomous vehicle using the control commands to cause the autonomous vehicle to maneuver along the 3D trajectory.
13. The method of claim 10 , wherein generating the cost function map comprises:
feeding the image into a machine learning model to determine the level of risk associated with navigating in the area of the physical environment corresponding to the region of the image.
14. The method of claim 10 , further comprising:
determining a cost value associated with each of the multiple regions of the image.
15. The method of claim 14 , wherein determining the cost value associated with a particular region of the multiple regions includes:
estimating depth to a physical surface in an area of the physical environment corresponding to the particular region of the image; and
determining a level of confidence in the estimated depth.
16. The method of claim 10 , further comprising:
receiving an additional image of the physical environment while the vehicle is in motion through the physical environment;
processing additional image to dynamically update the cost function map.
17. The method of claim 16 , further comprising:
identifying, based on the updated cost function map, a particular region of the additional image that is associated with a higher cost value than one or more of the other regions of the additional image; and
optimizing the 3D trajectory for an autonomous vehicle through the physical environment to lessen the overall cost of the 3D trajectory by avoiding the particular region of the additional image.
18. A vehicle comprising:
an image sensor configured to capture images of a physical environment;
a propulsion system configured to maneuver the vehicle through the physical environment; and
a visual navigation system coupled with the image sensor and the propulsion system, the visual navigation system configured to:
generate a cost function map that associates a cost value with each of multiple regions of an image of the physical environment,
wherein the cost value is indicative of a level of risk associated with navigating in an area of the physical environment corresponding to the region of the image; and
optimizing, using the cost function map, a 3D trajectory for an autonomous vehicle through the physical environment by reducing an overall cost of the 3D trajectory.
19. The vehicle of claim 18 , wherein the visual navigation system is further configured to:
identify, based on the cost function map, a particular region of the image that is associated with a higher cost value than one or more other multiple regions of the image;
wherein the 3D trajectory for the autonomous vehicle is optimized by adjusting the 3D trajectory such that a projection of the 3D trajectory in an image space of the image avoids the particular region of the image.
20. The vehicle of claim 18 , further comprising:
determining a cost value associated with each of the multiple regions of the image;
wherein determining the cost value associated with a particular region of the multiple regions includes:
estimating depth to a physical surface in an area of the physical environment corresponding to the particular region of the image; and
determining a level of confidence in the estimated depth.