The shortage of nurses who assist operating surgeons exchange surgical instruments has been chronically severe. To compensate for this shortage, we have been proposing and developed three versions of the Scrub Nurse Robot (SNR) system that is capable of functioning as a skilled human scrub nurse in endoscopic and laparoscopic surgery. In this paper, we mentioned those three versions of SNR and its features. At first, we developed 1st SNR, which was composed of two separated robots called R-1and R-2. The R-1 was designed to have two arms and it was aimed at handling only an electric knife and a blood aspirator which are the most frequently used during endoscopic and laparoscopic surgery. The R-2 was designed to have a horizontal arm, and two hands were mounted at the both ends of the arm. R-2 also has a tool changer storing instruments. It treats the other instruments than an electric knife and a blood aspirator. After the basic evaluation of lst SNR, we developed the 2nd version of SNR, and achieved smooth and wide movement of its arms each with 4 D.O.F.. The 2nd SNR is able to speak several sentences and recognize some words as well as the names of surgical instruments, and is also capable of recognizing a surgeon's intraoperative actions by its real-time visual recognition system (RTVRS). The RTVRS is basically composed of both a commercially-available 3D position tracking system and the algorithm that we developed to recognize surgeon's actions during exchange of instruments from the above-mentioned positional data. Concerning the results from evaluation of 2nd SNR, we developed the 3rd version of SNR which is able to insert the tip of instrument into a small-caliber trocar cannula penetrating a patient's abdominal wall. This function is named Automatic Instrument Inserting Function (AIIF). The AUF's algorithm termed "Kinematic based feedback control" enables the arm to insert the instrument into the cannula instead of a surgeon. We evaluated how quickly the 3rd SNR helped surgeon's stand-ins to exchange instruments in a laboratory, in comparison with human scrub nurses in real surgical operations and with the 2nd SNR that we have developed as the previous version. We defined the time for spending exchanging as 'targeting time'. The result of the evaluation showed that it was possible to exchange instruments 1.69 sec faster than that in clinical operating room between a human scrub nurse and a surgeon and 2.8 sec faster than that of the 2nd SNR.