Mobile body control device, mobile body control program, and mobile body

A mobile body control device that controls a moving unit of a mobile body such that the mobile body moves accompanying a subject at a position ahead of the subject in a traveling direction thereof, the control device comprising:

• a processor programmed to:
• acquire subject information including the position and traveling direction of the subject;
• process the subject information in time-series order to acquire, as an observation trajectory, a trajectory along which the subject has walked until a current time point;
• acquire, from a plurality of basic trajectories which are trajectories along which a walker walks toward a plurality of different target points, each of the basic trajectories being added, along its trajectory, with an uncertainty that represents probability distribution deviating from the trajectory, the basic trajectory that matches the observation trajectory as a specific basic trajectory;
• acquire from the specific basic trajectory, as a subject predicted trajectory, a trajectory along which the subject walks after the current time point so as to follow the observation trajectory;
• calculate, based on the uncertainty added along the subject predicted trajectory, prediction reliability of prediction that the subject walks following the subject predicted trajectory along the subject predicted trajectory;
• acquire, as a target trajectory, a trajectory formed by a target position of the mobile body set ahead of the subject in the traveling direction when it is assumed that the subject walks following the subject predicted trajectory;
• generate a control command sequence using an evaluation function that uses as input the subject predicted trajectory, prediction reliability, target trajectory, and a control trajectory obtained when the mobile body moves according to the control command sequence in time-series order, the evaluation function adding a control penalty term for the control command sequence, a subject penalty term based on a distance between the subject predicted trajectory and control trajectory, and a target penalty term based on a distance between the target trajectory and control trajectory and multiplies the prediction reliability as a coefficient with the subject penalty term and target penalty term to calculate the evaluation value, the control command sequence being generated so as to minimize the evaluation value;
• and
• control the moving unit based on the control command sequence.