A simple realtime software architecture has been devised to realise the manual guidance control. The architecture is composed of the following modules: a real-time UDP client that communicates with the C4Gopen server and executes the control algorithm, one or more real-time threads interfacing the client to the daq boards, and one or more non real-time tasks executing non real-time operations (e.g. logging of data), being connected to the client through real-time mailboxes.

A feasibility study of the vision system required to measure burr features has been performed. The aim of this study was to verify the availability of commercial hardware (a laser stripe and a digital camera) at a reasonable cost (at least compared to the cost of the overall system) that can achieve a scan resolution of0.1 mm. A camera and a laser stripe have been selected and will be soon tested.

Concerning the safety aspects during the walk-through programming phase, the C4G safety chain was modified in order to connect another safety device.

C4Gopen controller modalities have been studied in order to select a suitable one for the implementation of the manual guidance. Modality 5 has been selected.

Direct kinematics and direct/inverse differential kinematics were coded in the realtime PC programming environment.

The ATI functions were ported to Linux and tested in the real-time environment. The code that handles the acquisition board has been also improved, in order to speed up the acquisition of multiple channels.

To setup a manual guidance programming mode that does not violate the safety standards, the open controller should be active in programming mode, and the human operator should keep the dead-man button pressed.

An additional lighter and ergonomic dead-man button (like e.g. the Sick E100 enabling switch) will be added to the safety chain and the software of the C4Gopen will be changed in order to allow the activation of the open controller even in programming mode.

The preparation of the experimental setup was completed.
The Smart Six robot was installed, the external PC for the open controller was set up, the force/torque sensor was delivered.

The FIDELIO project wants to investigate the feasibility of an innovative robot programming approach, based on the Learning From Observation paradigm, in an industrial application scenario exemplified by a fixtureless wheel deburring task. To this aim, a robotised cell is being set up, composed of a Comau Smart-Six manipulator, equipped with a force-torque sensor, a camera and a pneumatic deburring tool, and a workstation where aluminium wheels are placed for deburring.