Home > Project Activity > Completed project  
이동 로봇 자세 제어를 위한 전자영상 안정화 연구
 
Humanoid biped Robot researchers are interested in vision of robot for Human-Robot Interface. Therefore today's most of the humanoid biped robots use vision camera to recognize objects like human or avoid obstacles. But they don't use vision camera as a tilt angle sensor to control attitude of robot. Because vision sensor has some limitations to estimate tilt angle of robot. First, vision sensor need very long processing time to calculate angle. Therefore it generates low sampling frequency and estimation time delay. And also, vision sensor should compare two pictures to estimate tilt angle so that the Field of view of vision sensor is limited. To apply vision sensor to control attitude of biped robot, we should solve the limitations of vision sensor. To solve low sampling frequency and estimation time delay, we suggest fusion filter for vision sensor with gyro sensor. Gyro sensor have also disadvantage which is drift generated by DC offset. Fusion filter compensate drift of gyro sensor by vision sensor, and also compensate low sampling time and time delay of vision sensor by gyro sensor. specific procedure of fusin filter is shown below.


Fig 1. Fusion filter structure for multi sampling rate sensors

 

Fig 2. Integration of gyro signal to compensate estimation time delay

To solve limited FOV of vision sensor, we suggest stabilization control system of vision sensor. As vision sensor is stabilized all the time, then vision can estimate tilt angle all the time.

Fig 3. Stabilization control structure for vision sensor

We expect that the vision system which has high sampling frequency, no time delay and unlimited FOV can be applied to biped robot to control attitude of robot like human do.