Autonomous vehicle fleet model training and testing

A method of controlling a vehicle comprising:

• receiving an indication that a vehicle controller of the vehicle has or will have unallocated computational resources at a time period;
• based at least in part on receiving the indication, causing the vehicle controller to one or more of:
• train a target machine-learning model, creating a trained model;
• or
• test an experimental model;
• and
• control the vehicle according to an output of the trained model or the experimental model.

A non-transitory computer-readable medium having a set of instructions that, when executed, cause one or more processors to perform operations comprising:

• determining that a vehicle has unallocated computational resources during a time period;
• transmitting to the vehicle instructions to conduct a test mission during the time period and one or more of an experimental model or instructions to train a target model stored at the vehicle;
• receiving one or more of sensor data, an updated target model, or performance metrics from the vehicle that correspond to completion of at least a portion of the test mission;
• and
• determining that one or more of the updated target model or the experimental model changed performance of the vehicle.