Automated Experiment Design
AI Planning techniques for automated robotic experiments
What is the challenge that is being addressed?
The Automated Experiment Design Domain aims at improving efficiency and flexibility of an industrial scenario at Procter&Gamble in which robots are employed to support people in the development of automated quality tests, considering in particular testing laundry detergent soluble pouches.
The main goal of this use case is to demonstrate the effectiveness of AI planning technology integrated with robotics solutions in delivering a high number of robotics procedures for the various use-cases reducing manual operations and planning that currently limit flexibility and efficiency of the system. Moreover, we would like to empower non-robotics-experts (i.e., P&G lab users) to be able to customise, adapt, and change the course of the robotics procedures, using natural interfaces.
The specific use-case considered here is Quality testing of unit dose pouches. There are several quality test methods on laundry detergent soluble pouches, such as weight, dimensions, elasticity, strength and tightness of the pouch, and for statistical significance they need to be run on a high number of samples (thousands).
What is the AI solution the project plans to implement?
The problem is modelled using the Unified Planning (UP) framework through the specification of predicates and actions needed to describe the application domain according to the requirements of the real scenario. UP planning engines will be used to compute plans solving the problem and they will then be integrated together with specific TSBs for a complete implementation of the solution in the industrial setting. A simulated environment replicating most of the features of the real one is available for development, test and validation.