Modular autonomous drone
An autonomous quadcopter has four motors, each motor coupled to a corresponding propeller and a flight controller coupled to the four motors to provide input to the four motors to control flight. The autonomous quadcopter also has a plurality of cameras and an Artificial Intelligence (AI) controller coupled to the plurality of cameras to receive input from the plurality of cameras, determine a flightpath for the autonomous quadcopter according to the input from the plurality of cameras, and provide commands to the flight controller to direct the flight controller to follow the flightpath.
1. An autonomous quadcopter comprising:
a chassis;
four motors mounted on the chassis, each motor coupled to a corresponding propeller;
a flight controller coupled to the four motors to provide input to the four motors to control flight of the autonomous quadcopter;
a Radio Frequency (RF) communication circuit coupled to the flight controller, the RF communication circuit configured to receive external commands from a remote-control and provide the external commands to the flight controller to direct the flight controller to follow a remotely-selected flightpath;
a plurality of cameras; and
an Artificial Intelligence (AI) controller coupled to the plurality of cameras, the AI controller removably attached to the chassis, the AI controller configured to receive input from the plurality of cameras, determine an AI-selected flightpath for the autonomous quadcopter according to the input from the plurality of cameras, and the AI controller connected to the flight controller to provide AI generated commands to the flight controller to direct the flight controller to follow the AI-selected flightpath, the AI generated commands and the external commands belong to a common command set, the AI controller is connected to the flight controller through a physical connector such that the AI controller is physically removable from the autonomous quadcopter to reconfigure the autonomous quadcopter for remote control according to the external commands, the physical connector is compatible with a simulator to enable connection of the AI controller to the simulator.
2. The autonomous quadcopter of claim 1 wherein the plurality of cameras includes three pairs of cameras, each pair of cameras configured to obtain a stereoscopic view of a different area around the autonomous quadcopter.
3. The autonomous quadcopter of claim 1 wherein the AI controller includes AI code to be trained by machine learning and a pre-installed machine learning library.
4. The autonomous quadcopter of claim 1 wherein the autonomous quadcopter is configurable to follow either the external commands from the remote-control or the AI generated commands from the AI controller when the AI controller is connected to the connector according to selection by a remote-control user.
5. The autonomous quadcopter of claim 1 further comprising at least one of a clip or bolt for physically attaching the AI controller to the chassis.
6. The autonomous quadcopter of claim 1 further comprising one or more of: a gyroscopic sensor, accelerometer, a range sensor, coupled to provide input to the AI controller.
7. The autonomous quadcopter of claim 1 further comprising a hardware abstraction layer configured to separate interleaved camera output received from cameras of the plurality of cameras and to separately buffer deinterleaved output such that frames of an individual camera are separately buffered and available at a predetermined frame rate.
8. The autonomous quadcopter of claim 1 wherein the AI controller includes an AI master, scheduler, and libraries of routines provided by a Software Development Kit (SDK) provider and includes participant code generated by a participant using the SDK.
9. An autonomous quadcopter comprising:
a chassis;
four motors attached to the chassis, each motor coupled to a corresponding fixed-pitch propeller;
a flight controller coupled to the four motors to provide input to the four motors to control flight of the autonomous quadcopter;
a Radio Frequency (RF) communication circuit coupled to the flight controller, the RF communication circuit configured to receive external commands from a remote-control and provide the external commands to the flight controller to direct the flight controller to follow a remotely-selected flightpath in a remote-control mode;
a plurality of stereoscopic cameras;
an Artificial Intelligence (AI) controller removably attached to the chassis, the AI controller coupled to the plurality of stereoscopic cameras and the flight controller, the AI controller configured to receive input from the plurality of stereoscopic cameras, perform hardware abstraction on the input from the plurality of stereoscopic cameras, analyze abstracted input from the plurality of stereoscopic cameras to determine locations of objects around the autonomous quadcopter, determine an AI-selected flight path for the autonomous quadcopter according to the locations of the objects, and provide AI generated commands to the flight controller to direct the flight controller to follow the AI-selected flight path in an autonomous mode, the AI generated commands and the external commands from the RF communication circuit defined by a common command set; and
a connector to connect the AI controller and the flight controller for operation in the autonomous mode with the AI controller located in the autonomous quadcopter, enable removal of the AI controller from the autonomous quadcopter for operation in the remote-control mode, and enable connection of the AI controller to a simulator.
10. The autonomous quadcopter of claim 9 wherein the common command set is a MAVLink command set.
11. The autonomous quadcopter of claim 9 wherein the plurality of stereoscopic cameras includes three stereoscopic cameras in different orientations with respect to the autonomous quadcopter.
12. The autonomous quadcopter of claim 11 further comprising one or more Inertial Measurement Unit (IMU) sensors coupled to the AI controller to provide autonomous quadcopter motion data to the AI controller.
13. The autonomous quadcopter of claim 12 further comprising a rangefinder coupled to the AI controller, the rangefinder configured to determine distance between an object and the autonomous quadcopter.
14. The autonomous quadcopter of claim 13 wherein the AI controller includes a hardware abstraction layer configured to receive input from the plurality of stereoscopic cameras, the one or more IMU sensors, and the rangefinder, and to buffer sensor and camera data.
15. The autonomous quadcopter of claim 14 wherein the hardware abstraction layer is further configured to provide output to a plurality of motors of the autonomous quadcopter.
16. A method of operating a quadcopter comprising:
generating a plurality of stereoscopic camera views for a plurality of fields of view around the quadcopter;
providing the plurality of stereoscopic camera views to a hardware abstraction layer;
deinterleaving of the plurality of stereoscopic camera views in the hardware abstraction layer;
providing deinterleaved frames to an Artificial Intelligence (AI) module that is removably attached to the quadcopter;
generating object-location information for objects in the plurality of fields of view from the deinterleaved frames in the AI module;
determining in the AI module an AI-selected flight path for the quadcopter according to the object-location information;
generating a plurality of AI-generated commands in the AI module corresponding to the AI-selected flight path;
sending the plurality of AI-generated commands from the AI module through a quadcopter connector to a flight controller of the quadcopter;
controlling a plurality of motors of the quadcopter according to the plurality of AI-generated commands such that the quadcopter follows the flight path;
disconnecting the AI module from the quadcopter connector;
removing the AI module from the quadcopter;
connecting the AI module to a simulator through a simulator connector;
while the AI module is connected to the simulator, testing AI code of the AI module;
while the AI module is disconnected and removed from the quadcopter, receiving by the quadcopter a plurality of remotely-generated commands, the remotely-generated commands and the AI-generated commands from a common command set;
sending the plurality of remotely-generated commands to the flight controller; and
controlling the plurality of motors of the quadcopter according to the plurality of remotely-generated commands such that the quadcopter without the AI controller follows a remotely-selected flight path.
17. The method of claim 16 wherein the objects in the plurality of fields of view include other quadcopters and racecourse features, and wherein determining the flight path for the quadcopter includes determining the fastest flightpath to complete a racecourse without hitting the other quadcopters or racecourse features.
18. The method of claim 17 wherein determining the flight path for the quadcopter occurs autonomously, without input from a human pilot.
19. The method of claim 17 wherein the plurality of remotely-generated commands are sent to the flight controller over an interface that is coupled to a Radio Frequency (RF) receiver to alternatively receive the remotely generated commands from the RF receiver according to the common command set.
20. The method of claim 19 further comprising, while the AI controller is physically connected to the quadcopter connector and attached to the quadcopter, disabling commands from the AI controller and enabling commands from the RF receiver in response to a command received by the RF receiver from a remote-control.