The article presents a logic branching weighted algorithm (LBWA) to train a robot to perform splined shaft and hole assembly in a robotic cell. The LBWA uses angular and linear positional changes and assigns weights to each of these based on the force sensing information from an assembly path and evolves a best move strategy for the robot to complete the task. The machine learning capability of the robot depends on the discretization of the force-torque information that is monitored and mapped for each position. Prior to commencing the move, the LBWA compares the evaluating functions. A trade-off is to be made between the information space and the learning time for the robot in a real-life situation. Experimental results are presented to establish the effectiveness of the LBWA in training the robot.