Travel route generation method for autonomous vehicle and control apparatus for autonomous vehicle

A control apparatus for an autonomous vehicle, the apparatus comprising:

• a processor configured to
• detect a boundary of a travel path based on scene information that includes the travel path in relation to forward of the autonomous vehicle and is obtained by a sensor for one frame at a predetermined sampling cycle,
• generate, based on the boundary detected, a provisional travel route for the autonomous vehicle on the travel path, and
• use a likelihood based on machine learning to evaluate the boundary,
• wherein
• the processor generates the provisional travel route based on the boundary such that a safety margin is formed between the provisional travel route and the boundary, and sets a size of the safety margin in accordance with the likelihood for the boundary,
• for a plurality of the provisional travel routes generated based on the boundary detected based on the scene information for each one frame, the processor determines whether or not there is a deviation greater than or less than a preset range for amounts of deviation, between the provisional travel route newly generated based on the scene information for one frame and the provisional travel route generated based on the scene information for another one frame previous to the one frame,
• in a case where the deviation is present, the processor updates the provisional travel route to the provisional travel route newly generated, and
• in a case where the deviation is not present, the processor does not update the provisional travel route to the provisional travel route newly generated.