Apparatus for diagnosing abnormality in vehicle sensor and method thereof

An apparatus for diagnosing an abnormality of a vehicle sensor, the apparatus comprising:

• a sensor configured to measure an acceleration and an angular velocity of a vehicle;
• a camera configured to generate a front time series image frame of the vehicle;
• and
• a controller configured to estimate the acceleration and the angular velocity of the vehicle by using the front time series image frame generated by the camera and to diagnose an abnormality in the sensor based on the acceleration and the angular velocity of the vehicle estimated by the controller, wherein the controller comprises:
• a pose network configured to predict first pose information of the camera by using an image at a time point of t−1 and an image at a time point of t and second pose information of the camera by using an image at a time point of t−2 and the image at the time point of t−1;
• a 6D estimation network configured to estimate the acceleration and the angular velocity of the vehicle by using the first pose information and the second pose information;
• and
• a diagnostic device configured to compare the acceleration and the angular velocity of the vehicle estimated by the 6D estimation network with the acceleration and the angular velocity of the vehicle measured by the sensor to diagnose the abnormality in the sensor, wherein the diagnostic device is configured to diagnose that the abnormality occurs in an acceleration sensor in the sensor when an error of the acceleration measured by the acceleration sensor is not included in a first error range based on the acceleration estimated by the 6D estimation network.

A method of diagnosing an abnormality of a vehicle sensor, the method comprising:

• measuring, by a sensor, an acceleration and an angular velocity of a vehicle;
• generating, by a camera, a front time series image frame of the vehicle;
• and
• estimating the acceleration and the angular velocity of the vehicle by using the front time series image frame generated by the camera and diagnosing whether there is an abnormality in the sensor based on the acceleration and the angular velocity of the vehicle estimated by the controller, wherein diagnosing whether there is an abnormality in the sensor comprises:
• predicting first pose information of the camera by using an image at a time point of t−1 and an image at a time point of t;
• predicting second pose information of the camera by using an image at a time point of t−2 and the image at the time point of t−1;
• estimating the acceleration and the angular velocity of the vehicle by using the predicted first pose information and the predicted second pose information;
• determining whether there is an abnormality in an acceleration sensor by comparing the estimated acceleration with the measured acceleration;
• and
• determining whether there is an abnormality in a gyro sensor by comparing the estimated angular velocity with the measured angular velocity;
• wherein determining whether there is the abnormality in the acceleration sensor includes:
• determining whether an error of the measured acceleration is included in a first error range based on the estimated acceleration;
• diagnosing that an abnormality occurs in the acceleration sensor when the error of the measured acceleration is not included in the first error range, as a determination result;
• and
• diagnosing that the abnormality does not occur in the acceleration sensor when the error of the measured acceleration is included in the first error range, as the determination result.