This paper presents a method for the generation of robotic assembly. By using mating condition classification as a medium, the problems of robotic assembly planning have been reconsidered. Two matrices such as interference matrix and mating matrix are used and an algorithm is developed for the generation of all feasible disassembly sequences. By reversing the disassembly sequences, all feasible robotic assembly sequences are thus generated. Finally, dynamic programming is applied for finding the best assembly sequence.