Typically, the instructions of a particular chemical experiment are described in natural language. Our autonomous mobile robot relies on a couple of mechanisms for performing an experiment from such a description. Initially, it infers the meaning of each natural language instruction to be performed and uses the background knowledge to chose and parametrize the most appropriate plan from the plan library for each given instruction. Then each chosen plan queries the vision system for the objects to be manipulated and runs the constraint motion controller to perform the necessary motions. Finally the logging mechanism records all details concerning how was the chemical experiment performed.
The logs could be used to answer a wide range of questions concerning the performed experiment, improving the way future experiments can be performed.
This project has received funding from ACAT and RoboHow: