Autonomous driving system, autonomous driving method, and autonomous driving program

1. An autonomous driving system comprising:

a recognition unit configured to recognize obstacles based on image data obtained by imaging a predetermined region including a road on which an autonomous vehicle travels, the imaging being performed by an imaging device installed at a specific location in the external environment of the autonomous vehicle;

a creation unit configured to create a feature quantity distribution expressing a distribution of features quantities relating to obstacles that have been recognized by recognition unit in the past for each of the locations where the obstacles were recognized and for each of the types of the obstacles; and

a judgement unit configured to compare (i) a feature amount distribution corresponding to the location and type of an obstacle recognized as an evaluation target created by the creation unit with (ii) a feature quantity of the obstacle recognized as the evaluation subject, to thereby judge whether the obstacle recognized as the evaluation subject is erroneously recognized.

2. The autonomous driving system according to claim 1 , wherein

the creation unit is configured to create, as the feature quantity distribution, at least one of a dwell time distribution expressing a distribution of dwell times of obstacles, a travel speed distribution expressing a distribution of travel speeds of obstacles, and a travel distance distribution expressing a distribution of travel distances of obstacles, and

the judgement unit is configured to compare at least one of the dwell time distribution, the travel speed distribution, and the travel distance distribution respectively created by the creation unit, with at least one of a dwell time, a travel speed, and a travel distance of the obstacle recognized as the evaluation subject, to thereby judge whether the obstacle recognized as the evaluation subject is erroneously recognized.

3. The autonomous driving system according to claim 2 , wherein the judgement unit is configured to perform the comparison based on at least one of a representative value obtained from a statistical analysis of the dwell time distribution, a representative value obtained from a statistical analysis of the travel speed distribution, and a representative value obtained from a statistical analysis of the travel distance distribution.

4. The autonomous driving system according to claim 2 , wherein the judgement unit is configured to perform the comparison based on at least one of a threshold value obtained from the dwell time distribution, a threshold value obtained from the travel speed distribution, and a threshold value obtained from the travel distance distribution.

5. The autonomous driving system according to claim 4 , further comprising a notification unit configured to, in response to the judgement unit judging that an obstacle is erroneously recognized, transmit obstacle information to an autonomous vehicle traveling in the surroundings of the obstacle, the obstacle information having a flag attached for removing the recognition results for the obstacle.

6. The autonomous driving system according to claim 5 , further comprising a presentation unit configured to, in response to the judgement unit judging that an obstacle is erroneously recognized, present image data to an operator that corresponds to the time at which the judgement of erroneous recognition occurred, and receive, as input from the operator, an indication as to whether there is an anomaly, and

wherein the notification unit is configured to, in response to an input notifying an anomaly being received from the operator, send the obstacle information to an autonomous vehicle traveling in the surroundings of the obstacle, the obstacle information having a flag attached for removing the recognition results for the obstacle.

7. The autonomous driving system according to claim 1 , further comprising a learning unit configured to, in response to an obstacle being judged by the judgement unit to be erroneously recognized, perform learning of an obstacle recognition model, the learning being performed by using judgement results input by the operator as a label, and the judgement results including an indication of whether there is an anomaly and the causes thereof.

8. An autonomous driving method comprising:

recognizing obstacles based on image data obtained by imaging a predetermined region including a road on which an autonomous vehicle travels, the imaging being performed by an imaging device installed at a specific location in the external environment of the autonomous vehicle;

creating a feature quantity distribution expressing a distribution of feature quantities relating to obstacles that have been recognized in the past for each of the locations where the obstacles were recognized and for each of the types of the obstacles; and

comparing (i) the created feature amount distribution corresponding to the location and type of an obstacle recognized as an evaluation target with (ii) a feature quantity of the obstacle recognized as the evaluation subject, to thereby judge whether the obstacle recognized as the evaluation subject is erroneously recognized.

9. A non-transitory computer-readable storage medium for storing an autonomous driving program for causing a computer to execute a process, the process comprising:

recognizing obstacles based on image data obtained by imaging a predetermined region including a road on which an autonomous vehicle travels, the imaging being performed by an imaging device installed at a specific location in the external environment of the autonomous vehicle;

creating a feature quantity distribution expressing a distribution of features quantities relating to obstacles that have been recognized in the past for each of the locations where the obstacles were recognized and for each of the types of the obstacles; and

comparing (i) a feature amount distribution corresponding to the location and type of an obstacle recognized as an evaluation target created with (ii) a feature quantity of the obstacle recognized as the evaluation subject, to thereby judge whether the obstacle recognized as the evaluation subject is erroneously recognized.