Objective-based control of an autonomous unmanned aerial vehicle
Techniques are described for controlling an autonomous vehicle (AV) using objective-based inputs. The underlying functionality of an autonomous navigation system can be is exposed via an application programming interface (API) or similar methodology allowing the UAV to be controlled through specifying a behavioral objective, for example, using a call to the API to set parameters for the behavioral objective. The autonomous navigation system can then incorporate perception inputs such as sensor data from sensors mounted to the UAV and the set parameters using a multi-objective motion planning process to generate a proposed trajectory that most closely satisfies the behavioral objective in view of certain constraints. Developers can utilize the API to build customized applications for the AV. Such applications, also referred to as “skills,” can be developed, shared, and executed to control behavior of an autonomous AV.
1 . A method for autonomous control of an unmanned aerial vehicle (UAV) through a physical environment using behavioral objectives defined via a navigation application programming interface (API), the method comprising:
receiving, by a computer system, sensor data from a sensor onboard the UAV;
receiving, at the API, information indicative of a behavioral objective;
exposing, via a public-facing navigation API of the UAV, parameterizations of the behavioral objective to a third-party application;
receiving, at the API from the third-party application, a call that sets or modifies one or more parameters of the behavioral objective, the parameters including at least one of: a target, a dead-zone region about the target, or a weighting factor;
inputting, by the computer system, the sensor data and the one or more parameters into a multi-objective trajectory generation process to generate a proposed trajectory that most closely satisfies the behavioral objective in view of another behavioral objective or constraint;
enforcing, by the computing system, one or more built-in safety objectives including obstacle avoidance and airframe dynamic limits as immutable constraints that are not alterable via the API; and
generating, by the computer system, control commands configured to cause the UAV to autonomously maneuver through the physical environment based on the proposed trajectory.
2 . The method of claim 1 , wherein the behavioral objective is defined relative to a semantic understanding of the physical environment and requires maintaining saliency of a specified object class or activity in captured images, including avoiding backlighting of a tracked subject.
3 . The method of claim 1 , wherein the multi-objective trajectory generation is performed continually as updated sensor data and/or updated parameters of the behavioral objective are received, and wherein API-provided parameter updates are validated against one or more immutable safety objectives, any parameter update that would violate a safety objective is rejected, and each rejected update is recorded in an audit log.
4 . The method of claim 1 , wherein the dead-zone region defines a tolerance region around the target within which the multi-objective trajectory generation refrains from adjusting the proposed trajectory.
5 . The method of claim 1 , wherein the multi-objective trajectory generation is performed continually as updated sensor data and/or updated parameters are received.
6 . The method of claim 1 , wherein the weighting factor defines a relative influence of the behavioral objective with respect to another behavioral objective or constraint.
7 . The method of claim 1 , further comprising logging the parameters received from the third-party application and a hash of the generated trajectory to an audit record.
8 . The method of claim 1 , wherein the built-in safety objectives include collision-avoidance and conformance with vehicle dynamic limitations.
9 . The method of claim 1 , wherein the multi-objective trajectory generation selects among candidate trajectories by minimizing a weighted cost function of the behavioral objective and the built-in safety objectives.
10 . An unmanned aerial vehicle (UAV) configured for autonomous flight through a physical environment, the UAV comprising:
an image capture device; and
a navigation system including an application programming interface (API), the navigation system configured to:
expose information indicative of a behavioral objective;
receive information indicative of a behavioral objective;
parameterize the behavioral objective by receiving, from a third-party application, a call that sets or modifies one or more parameters of the behavioral objective, the parameters including at least one of: a target, a dead-zone region about the target, or a weighting factor;
input sensor data and the one or more parameters into a multi-objective trajectory generation process to generate a proposed trajectory that most closely satisfies the behavioral objective in view of another behavioral objective or constraint;
enforce one or more built-in safety objectives including obstacle avoidance and airframe dynamic limits as immutable constraints that are not alterable via the API; and
generate control commands configured to cause the UAV to autonomously maneuver through the physical environment based on the proposed trajectory.
11 . The UAV of claim 10 , wherein the dead-zone region defines a tolerance region around the target within which the multi-objective trajectory generation refrains from adjusting the proposed trajectory.
12 . The UAV of claim 10 , wherein the weighting factor defines a relative influence of the behavioral objective with respect to another behavioral objective or constraint.
13 . The UAV of claim 10 , wherein the built-in safety objectives include collision-avoidance and conformance with vehicle dynamic limitations.
14 . The UAV of claim 10 , further configured to continually perform multi-objective trajectory generation as updated sensor data and/or updated parameters are received.
15 . The UAV of claim 10 , wherein the API receives and the UAV persists a record of the parameter values and associates the record with a corresponding flight log.
16 . An apparatus comprising:
one or more non-transitory storage media; and
instructions stored on the one or more non-transitory storage media that, when executed by one or more processors, cause the one or more processors to:
receive sensor data from a sensor onboard an unmanned aerial vehicle (UAV);
receive information indicative of a behavioral objective;
expose parameterizations of a behavioral objective to a third-party application;
receive a call that sets or modifies one or more parameters of the behavioral objective, the parameters including at least one of: a target, a dead-zone region about the target, or a weighting factor;
input the sensor data and the one or more parameters into a multi-objective trajectory generation process to generate a proposed trajectory that most closely satisfies the behavioral objective in view of another behavioral objective or constraint;
enforce one or more built-in safety objectives including obstacle avoidance and airframe dynamic limits as immutable constraints that are not alterable via the third-party application; and
generate control commands configured to cause the UAV to autonomously maneuver through the physical environment based on the proposed trajectory.
17 . The apparatus of claim 16 , wherein the dead-zone region defines a tolerance region around the target within which the multi-objective trajectory generation refrains from adjusting the proposed trajectory.
18 . The apparatus of claim 16 , wherein the multi-objective trajectory generation is performed continually as updated sensor data and/or updated parameters are received.
19 . The apparatus of claim 16 , wherein the weighting factor defines a relative influence of the behavioral objective with respect to another behavioral objective or constraint.