Brief description of functionality/utility:
ERL researchers have developed algorithms that allow a human operator to remotely control a robot as it autonomously
cuts through contaminated metal. The operator uses a computer mouse to control the start and end points of the cutting
tool path, with decisions being based on a 3D image of the object that has been captured by the robot and transmitted on
a screen. The algorithm works even on metal objects that have an arbitrary shape
The cutting path can be automatically optimised to avoid obstacles on the surface. It can also be programmed so that
disturbances or emergencies can be handled fast.
- The new algorithms work regardless of the type of robot or cutting tool adopted for the task.
- TRL 7 has been demonstrated in a live radioactive hot-cell environment at the Springfields nuclear site - This is
regarded internationally as currently world-leading in this domain. The autonomous cutting in an active cell
remains unrepeated worldwide since our landmark demo with UK National Nuclear Lab.
Decommissioning involves many cutting operations, in which arbitrarily shaped legacy structures and cell furniture must
be dismantled. Conventional programming of robot motions involves standing near the robot while teaching a large
number of tool positions which are then used as waypoints. This is only feasible in (say) manufacturing where exactly
the same product is being assembled many times. It is not scalable in decommissioning scenarios where a new object,
with new surface shape, may be encountered in each operation.
The Birmingham ERL team has demonstrated how:
Object surfaces, and their positions relative to the robot, can be accurately captured and modelled by advanced 3D
computer vision systems. Advanced algorithms can automatically plan:
- A correct trajectory for the cutting tool over the surface, maintaining the correct stand-off distance and orientation
to the surface.
- Trajectories for all joints of the robot, to deliver the tool along the cutting path, while avoiding obstacles and
problematic robot configurations.