Underactuated Robot Manipulators

Underactuated robots are the robots with both active and passive joints. For example, when a regular fully-actuated robots has failed joint motors, one wants to be still able to control the system if possible. For an hyper-redundant (e.g., more than 30 DOF), snake-like, robot, one would like to see if it works when some of joints are not actively controlled. In space applications, the free-flying base where a robot is mounted can be considered as a linkage with 6 passive virtual joints, and the overall system can be considered as an underactuated robot. Think about the martial-art superstar Bruce Lee who used to play a pair of numbchucks (three-link-sticks) as his favorite weapon. Each numbchuck is composed of two passive joints connecting three hardwood segments, and Bruce could fast project the tip of each numbchuck to a target, producing a tremendous impact force acting on the target. If we consider the numbchuck as a linkage with passive joints and Bruce's arm as a linkage with active joints, the overall combination is a typical underactuated system.

The issues that we are interested in investigating include:

We have been studying the above issues since 1994, and have accomplished substantial results. We have defined the dynamic coupling index to measure the amount of dynamic coupling between the actuated and the passive joints of an underactuated manipulator. We have developed the robust control schemes to stabilize all joints of an underactuated manipulator to an equilibrium point. We have showed, based on results from differential geometry, that a non-zero coupling index implies controllability of the passive joints via application of torques at the actuators.

We built two underactuated manipulators, one with 2 links and 1 passive joint, and another with 3 links and 2 passive joints. We use the brakes in our control method to eliminate the nonholonomic constraints in the dynamic equations associated with the passive joints. Both robots are locally controllable anywhere inside their workspaces, since their coupling indices are both non-zero at any given configuration. Control of all joints is then possible with only one actuator.

We will be looking into the following research issues in the future:

Researchers in this project are:

Publications in this project include: