Performance of autonomous vehicle operation in varying conditions by using imagery generated with machine learning for simulations

A non-transitory computer-readable medium having logic stored thereon that, in response to execution by a computing system, causes the computing system to perform actions for simulating an unmanned aerial vehicle (UAV) to improve control system performance, the actions comprising:

• obtaining, by the computing system, ground truth aerial imagery for a region, wherein the ground truth aerial imagery depicts the region during a first state;
• determining, by the computing system, a route for a simulated UAV within the region;
• generating, by the computing system based on the ground truth aerial imagery, predicted aerial imagery that depicts portions of the region associated with the route during the first state;
• generating, by the computing system, simulated aerial imagery that depicts portions of the region associated with the route during a second state different from the first state by providing the predicted aerial imagery to a machine learning model;
• simulating, by the computing system, travel of the simulated UAV along the route during the second state by providing the simulated aerial imagery as simulated input to at least one control system of the simulated UAV, and determining a simulated autonomous control decision generated by the at least one control system based on a given portion of the simulated aerial imagery during the simulated travel during the second state, wherein the simulated autonomous control decision generated by the at least one control system includes one or more of a landing location, a delivery location, a position, a path, or a command to an actuator or propulsion device;
• simulating, by the computing system, travel of the simulated UAV along the route during the first state by providing the predicted aerial imagery as simulated input to the at least one control system of the simulated UAV, and determining a ground truth autonomous control decision generated by the at least one control system based on a portion of the predicted aerial imagery corresponding to the given portion of the simulated aerial imagery during the simulated travel during the first state, wherein the ground truth autonomous control decision generated by the at least one control system includes one or more of a ground truth landing location, a ground truth delivery location, a ground truth position, a ground truth path, or a ground truth command to the actuator or propulsion device;
• and
• updating, by the computing system, the at least one control system based on the simulated autonomous control decision by comparing one or more of the landing location, the delivery location, the position, the path, or the command to the actuator or propulsion device of the simulated autonomous control decision to one or more of the ground truth landing location, the ground truth delivery location, the ground truth position, the ground truth path, or the ground truth command to the actuator or propulsion device of the ground truth autonomous control decision.