The Optimal Trajectory Control Of Tripod Robot Based On Visual Processing And Genetic Algorithm

In industrial production, objects often need to be picked up and placed in the desired location.The action is monotonous and repetitive. Operators are easily exhausted and inefficient. However,vision robot can finish the job easily, whose accuracy and rapidity is incomparable. In this paper,the process of industrial vision robot’s actual operation was simplified as the process of precisepick and place of geometries on a platform. The experimental research was conducted with aTripod robot system, which was designed by B&R Company.In order to improve the accuracy of pick and place, and increase the speed of this process,vision processing and path planning algorithm was designed in this paper. The experimentalresearch was divided into four modules, including image acquisition and preprocessing, centerlocation and color recognition of the platform and geometries, optimal grasping route and TCP/IPcommunication.Image acquisition and preprocessing module. A side view of the circular platform wasobtained with a camera fixed on the control cabinet in the proper place. The perspectivetransformation was used to correct the distortion image. Gray scale was applied to remove thecolor information and retain the luminance information of the image. Gaussian filtering wasutilized to eliminate the noise. A kind of improved local threshold OSTU method was designed forbinaryzation. At last, Canny operator edge extraction method was adopted to extract a satisfiedtarget contour.Center location and color recognition of the platform and geometries module. A two-steplocalization method was created to gain the placement location and the precise position ofplatform and geometries. Fast Hough transform was used to get the coarse center point of eachcircle. Coordinate rotation was adopted to gain the coarse center point of each triangle and star.The accurate positioning coordinate of each circle was estimated with vertical theorem. The accurate positioning coordinate of each triangle and star was determined by calculatingthe average value of pixels. After extracting RGB component value of each geometry on the outerring, the color recognition was realized by cvMat function.Optimal grasping route module. A method of genetic algorithm was designed to get theoptimal grasping trajectory scheme. To be specific, chromosomes were encoded, population wasinitialized, adaptive value was selected and genetic operators were chosen. Under the constraint ofputting geometries with same color together, the grasping sequence of the nine geometries wasobtained successfully. Meanwhile, the requirement of rapidity was satisfied.TCP/IP communication module. Through writing communication program in Visual Studioand setting up corresponding IP address and port, we achieved to transform coordinateinformation into message data with correct format, and transfer it to B&R PCC. After Real VNCreceived packet data successfully, Tripod robot could drive the mechanical arms to move throughits motion control system, which accomplished data processing and generated controlinstructions.The designed algorithm of visual processing and path planning was applied to the Tripodrobot platform. The robot arms picked up the geometries on the outer ring according to thepredetermined path, and placed them into the corresponding hole sites in the inner ring. Theaccuracy of grasping, correct rate of color recognition, precision of placement and processingrapidity were all satisfactory. Experimental results showed that the algorithm implemented on theTripod hardware platform had made its point. The specific purpose of pick and place was achievedaccurately and rapidly. This research has great value for expanding the application scope of Tripodrobot.