Session G112 - IBM Robotics Research Review

SHARE 70
February 29 - March 4, 1988

Gary Silverman (IBM) talked about current robot research. He brought a videotape that described some of the Yorktown Heights research projects, and described some practical applications.

Robot kinematics - physical construction - is well understood and well taught in the graduate schools, but a number of theoretical issues are still research topics. Among these are:

Uncertainty:
No measurement is exact, and no prediction will ever be 100% accurate. Robots have to contend with uncertain conditions.
Spatial reasoning:
What is the shortest path between the robot and an object? What if there is a box in the way? How does the robot determine how to walk around the box? What if the object is under another object? Robots have to be able to perform real physical navigation.
Compliance:
Robots have to adjust to continually changing conditions. "Compliant" robots adjust themselves to the environment or compensate for errors in measurement.
He showed us several pictures of walking robots. Researchers are very interested in developing walking robots, because when robots are bolted down they lose a considerable amount of flexibility. Some of the examples: Robots have to sense their environment. There are two types of sensing: "contact" and "non-contact". Contact sensors include touching, non-contact sensors are usually visual systems. A combination of the technologies was used to build a robot that sheared sheep! You strap the sheep down and turn the robot on; it actually ran with fewer errors than the human shearer, but was substantially slower.

The three videotapes Dr. Silverman showed talked about "camera calibration", "fine positioning" and "total hip replacement". Of the three, the last one was the most fascinating. He described a research project in which a robot will assist in a total hip replacement surgery (Silverman contended that this means that the field of Robotics was "totally hip". We booed him into submission.)

In hip replacement, the leg is opened and the "ball" part of the hip is sawed off the femur. A cavity is drilled into the femur, an implant (with a new ball) is inserted, and the femur reattached to the hip. The implant is usually secured in the femur cavity with a lot of cement; one of the problems with the technique is that the cement will only hold for 5-20 years, after which you have to do the surgery again.

An alternate, and preferred method of securing the implant is to allow the bone to attach itself to the implant. Doing this makes the implant last for the life of the patient, but the implant must match the cavity within close tolerances. Surgeons are typically unable to machine the cavity this exactly.

In the research project, sensors are attached to the patient's leg and a CAT scan is done. The CAT is fed to a computer, which builds an internal model of the patient's leg. The computer generates machining instructions for a custom implant to fit the patient's femur. When the surgery is done, the surgeon will secure the femur and activate a robot, which will drill the cavity to exactly fit the computer-designed implant.


Back to session index
Back to index of SHARE meetings
Read the disclaimer