Method, apparatus and program of predicting obstacle course
A method, an apparatus, and a program of predicting an obstacle course, capable of appropriately predicting a course of an obstacle even under a complicated traffic environment, are provided. For this purpose, the course, which the obstacle may take, is predicted based on the position and the internal state of the obstacle, and at the time of the prediction, a plurality of courses are probabilistically predicted for at least one obstacle. When there are a plurality of obstacles, the course in which different obstacles interfere with each other is obtained from the predicted courses, which a plurality of obstacles may take, and the predictive probability of the course for which the probabilistic prediction is performed from the courses in which they interfere with each other is lowered. Probability of realizing each of a plurality of courses including the course of which predicted probability is lowered is calculated.
1. An obstacle course prediction method, in which a computer predicts a course of an obstacle, the method comprising:
detecting, by a detector mounted on a vehicle, at least a plurality of positions and internal states of obstacles present within a predetermined area relative to the vehicle;
extracting a plurality of obstacles within the predetermined area based on the detected positions and internal states of obstacles within the predetermined area;
storing, in a non-transitory computer readable memory, at least the plurality of positions and internal states of the obstacles present within the predetermined area relative to the vehicle;
selecting a first obstacle from the plurality of obstacles;
calculating a probability of a plurality of first courses that the first obstacle may take, based on a position and an internal state of the first obstacle read from the memory;
selecting a second obstacle from the plurality of obstacles, the second obstacle being different from the first obstacle;
predicting a second course that the second obstacle may take, based on a position and an internal state of the second obstacle read from the memory;
determining whether or not each course in the plurality of first courses interferes with the second course;
in response to a determination of interference for a respective course in the plurality of first courses, lowering the probability of the respective course and modifying the probability of the plurality of first courses;
generating an operation signal based on the modified probability of the plurality of first courses; and
transmitting the operation signal to an actuator device provided in the vehicle.
2. The obstacle course prediction method according to claim 1 , further comprising:
obtaining a course, within the plurality of first courses, in which the first obstacle interferes with the second obstacle;
lowering the calculated probability of the obtained course;
calculating a probability of the first obstacle traveling along each of the plurality of first courses including the course in which the calculated probability of which is lowered due to the interference of the second obstacle, wherein
the first obstacle satisfies a predetermined condition;
calculating the probability of the plurality of first courses includes performing probabilistic prediction of a plurality of courses, which the first obstacle may take;
lowering the calculated probability of the obtained course includes lowering a predicted probability of a course of which distance at a same time from a course, which the second obstacle may take, is smaller than a predetermined value, within the plurality of courses, which the first obstacle may take, when there are a plurality of the obstacles, and
calculating a probability of realizing each of the plurality of first courses includes the course the predicted probability of which is lowered when lowering the calculated probability of the obtained course.
3. The obstacle course prediction method according to claim 2 , wherein the first obstacle course prediction includes:
trajectory generation including generating variation in position, which the first obstacle may take with time, as a trajectory in space-time composed of time and space, based on the position and the internal state of the first obstacle, and
performing probabilistic prediction calculation of a course, which the first obstacle may take, based on the trajectory generated at the trajectory generation.
4. The obstacle course prediction method according to claim 2 , wherein the second obstacle course prediction predicts the course of the second obstacle assuming that the internal state of the second obstacle is maintained.
5. The obstacle course prediction method according to claim 2 , further comprising:
generating a course of the vehicle based on a position and an internal state of the vehicle; and
calculating collision probability between the generated course of the vehicle and each of the plurality of courses, which the first obstacle may take, of which probability is calculated for each course at the probability calculation.
6. The obstacle course prediction method according to claim 1 , further comprising:
obtaining a course, within the plurality of first courses, in which the first obstacle interferes with the second obstacle;
lowering the calculated probability of the obtained course;
calculating a probability of the first obstacle traveling along each of the plurality of first courses including the course in which the calculated probability of which is lowered due to the interference of the second obstacle, wherein
calculating the probability of the plurality of first courses includes performing probabilistic prediction of the plurality of courses that the first obstacle may take,
lowering the calculated probability of the obtained course includes lowering a probability of taking a course of which distance at a same time from a course of another of the obstacles is smaller than a predetermined value within the courses, which the plurality of obstacles may take, when there are the plurality of obstacles, and
calculating a probability of realizing each of the plurality of first courses includes calculating the probability of realizing all the courses of the plurality of obstacles including the course the calculated probability of which is lowered when lowering the calculated probability of the obtained course.
7. The obstacle course prediction method according to claim 6 , wherein
calculating the probability of a plurality of first courses includes
trajectory generation including generating variation in position, which the first obstacle may take with time, as a trajectory in a space-time composed of time and space, based on the position and the internal state of the first obstacle, and
performing probabilistic prediction calculation of the course of the first obstacle based on the trajectory generated at the trajectory generating.
8. The obstacle course prediction method according to claim 6 , further comprising:
generating a course of the vehicle based on a position and an internal state of the vehicle; and
calculating collision probability between the course of the vehicle generated at the course generating and all the courses, which the first obstacle may take, probability of which is calculated for each course at the probability calculation.