taking　the　piston　blank　feeding　line　as　the　research　background,　motion　planning　for　grabbing　the　random　stacking　of　piston　blanks　is　studied.　Aiming　at　the　obstacle　avoidance　of　the　manipulator　caused　by　the　obstacle　piston　with　unknown　posture　around　the　target　piston,　an　improved　RRT　algorithm　is　proposed.　The　algorithm　divides　the　planning　process　into　two　phases　by　introducing　a　pre-pose　point.　In　the　first　stage,　the　main　purpose　is　to　avoid　the　collision　between　the　grippers　of　the　robotic　arm　and　the　obstructing　piston.　The　3D　mapping　of　area　around　the　target　piston,　and　the　gripping　direction　of　the　robotic　arm　is　obtained　based　on　the　principle　of　the　highest　safety　factor.　Afterward,　the　path　between　the　target　point　and　the　pre-posture　point　is　calculated　by　the　space　coordinate　transformation　formula.　The　main　purpose　of　the　second　phase　is　to　improve　the　path　quality　of　the　plan　under　the　premise　of　fully　exploiting　the　computational　efficiency　of　the　RRT　algorithm.　The　path　point　is　searched　using　the　mechanical　function　as　a　cost　function.　Based　on　the　shortest　path　principle,　the　path　from　the　pre-gesture　point　to　the　home　point　is　obtained　after　20　iterations.　Finally,　the　RRT　algorithm　of　this　paper　is　configured　in　the　ROS　environment　with　the　piston　blank　feeding　platform.　The　experimental　results　show　that　the　mechanical　arm　can　replace　the　person　to　complete　the　feeding　of　the　piston　blank.　©　2018　IEEE.