An efficient visual PID control approach for mobile robot

The ability of an autonomous mobile robot to control its distance and angle relative to other object is the basis of multi-robot or human/robot cooperation. Most of the approaches use costly sensor such as laser sensor or stereovision which often need complex calibration. This paper presents a novel uncalibration monocular visual PID approach for an autonomous robot to maintain its distance and angle relative to an object by detecting the object which has been trained through SIFT(Scale Invariant Feature Transform) algorithm offline using only an inexpensive video camera. The approach comprises two steps: (1) a visual detection unit which can provide information about the relative pose of the object by online DDM-BBF(Double Direction-Best Bin First) based SIFT feature matching. The relative distance and angle is detected by a least square solution of affine transform. (2) Two PID control loops are built on these visual inputs to adjust the robot's translational and rotational velocity which endows the robot with the ability of maintaining its distance and angle relative to a moving object. Experimental results proved the validity of the approach.