American astronauts aboard the international space station gave the outpost's new robotic arm a thorough workout on Thursday by successfully rehearsing the job of installing a 12-ton airlock during the next planned shuttle visit.

But the test started off with the arm's free end striking the station.

Expedition Two crewmembers Jim Voss and Susan Helms told ground controllers the arm's latching end effector, or LEE, suddenly jumped off its anchor only to rebound and hit the complex. They radioed hearing a thump.

File photo of astronauts Voss and Helms working at the controls of the Canadian-built space station robotic arm. Photo: NASA

Mission controllers in Houston speculated that one of the snare wires on the arm's end might not have fully retracted, allowing it snag the station's pin-like grapple fixture as the arm was moved away.

However, the astronauts said they weren't touching the controls for the arm at the time and that the crane moved from the berth by itself.

The $600 million arm appeared to survive unscathed. The astronauts performed a check of the latching end effector and didn't report any damage.

Later, controllers suspected that built up tension on the arm had caused it to spring away.

Despite the incident, the crew pressed ahead with the much-delayed demonstration test of movements the arm will make during next month's installation of the U.S.-made Joint Airlock.

The robotic arm will hoist the airlock from shuttle Atlantis' payload bay and mount it to the station's Unity node.

The complex maneuver can't be performed by the shuttle's arm because it isn't long enough to reach the airlock's docking port on the station.

The so-called "dry run" of the airlock installation has been on hold for several weeks due to a pair of problems experienced in earlier arm tests.

First, the arm's brakes engaged without being told to do so. Officials have since determined it was a one-time error because the problem has never repeated during dozens of tests.

The more significant trouble occurred when an intermittent communications problem cropped up between the arm's control computer and the Shoulder Pitch joint electronics.

The international space station's new robotic arm, seen here forming a V-shape. Photo: NASA

Since success of Atlantis' delivery mission hinges on the arm, NASA wants to ensure the 58-foot long crane is working properly before the shuttle and its five-person crew blasts off from Kennedy Space Center.

That launch was originally slated for Thursday, but the space agency has postponed the flight until July 12 at the earliest.

A series of computer software patches are being developed in an effort to diagnose the Shoulder Pitch joint trouble. But the communications glitch has not reappeared in about two weeks, stumping engineers trying to determine the mysterious cause of the problem.

When the communications are interrupted, the arm enters a safe mode and freezes in its position.

The problem has only occurred when the arm is running on its backup commanding system. The prime string has worked flawlessly.

Officials hope the arm's software could be altered in a way to "mask out" whatever the problem is. But the trouble must first be found.

Another option being developed would change the software so the backup control system wouldn't know the Shoulder Pitch joint even exists. That would mean the arm would operate with only six of its seven joints.

One option that has been ruled out is removing and replacing the Shoulder Pitch joint during the August flight of space shuttle Discovery.

NASA decided that if software changes couldn't correct the problem and replacing the joint was the only choice left, the massive job wouldn't be attempted until the November shuttle flight of Endeavour.

While the arm troubles are being worked out, the space shuttle launch schedule remains in limbo.

Space shuttle Discovery is moved from its hangar to the Vehicle Assembly Building earlier this week. Photo: NASA-KSC

Given the concerns about the arm's ability to perform the airlock installation, NASA has toyed with the idea of flip-flopping the next two shuttle launches -- Atlantis' airlock mission and the flight of Discovery to exchange the station's resident crews.

The situation is rather complicated.

Atlantis remains inside Kennedy Space Center's 52-story Vehicle Assembly Building awaiting rollout to launch pad 39B. That move could happen as soon as Tuesday.

The ship was transferred from its hangar to the VAB late last month to be mated with an external fuel tank and solid rocket boosters. But NASA put on hold the rollout to the pad until it becomes clear when Atlantis will fly.

Targeted for liftoff at 5:04 a.m. EDT July 12, NASA only has until July 17 to get the shuttle airborne. The launch can't occur between July 18 and August 3 because the solar heating on the shuttle docked to the station during that time would be too high for Atlantis' systems.

Discovery is scheduled for launch no earlier than 7:05 p.m. EDT on August 5 to ferry the Expedition Three crew to the station and return Expedition Two back to Earth, capping its five-month tour-of-duty.

Station officials want the crew rotation to occur in August. That means if Atlantis doesn't fly in July, it's mission is off until late September.

The Expedition Three astronauts, who will live aboard the station until early December, have started training in simulators to learn how to install the airlock using the station's arm if that job is delayed until their stay on the outpost.

Senior managers are slated to meet Monday to review the situation and determine if Atlantis can launch next month or if more time will be needed to diagnose the arm communications problem. If additional testing were needed, Discovery's flight would become next in line.

Discovery was moved into the VAB on Tuesday and has been attached to its tank and booster rockets. Rollout to launch pad 39A could occur by late next week.