 Robonaut removing an ORU on space station Freedom. Note the "stinger" tail is attached to a rail while the arms remove the ORU. Also note the hand rails on the space station’s superstructure. |
Robots have found a niche working in dangerous environments. Robots today work in deep sea operations, space missions, nuclear cleanup, and bomb disposal.
Because these robots are in situations that are hazardous for humans, a robot failure can be very expensive. In these failure critical missions, robotic systems must be fault tolerant.
The problem of failure recovery was first addressed in the aerospace community. In 1977, a pilot flying a Lockheed L-1011 landed the plane despite the complete lack of conventional pitch control [Montoya, 1982].
This was revolutionary because the pilot had used redundant control (in this case the engines) to control a plane that was out of control in the traditional sense. |
|
 A redundant robot demonstrating its self motion or null space. Note there are infinite manipulator configurations for a single end effector position. |
Using redundant serial robots for failure recovery seems simple, if an actuator fails, the controller locks the faulty joint and the redundant actuators continue operation. Unfortunately, this problem has proved to be deceptively difficult.
Following a joint failure, a robot may possess a sufficient number of actuators to address the end-effector space, but geometric singularities may leave the robot unable to work. In addition, locking a joint will change a robot's workspace.
The problem summary is:
Given an actuator or sensor experiences a failure, what is the impact of that failure and how should the manipulator be reconfigured in order to provide continued operation?
The problem of post failure recovery revolves around a principle condition - the effective use of the null space to continue operation, maintain positional control and provide the best overall performance. |
|
