Working in the Computer-Human Interaction in Learning and and Instruction, in this thesis, we are giving the first steps towards equipping social robots with state-of-the-art Visual Speech Recognition (VSR) for in-the-wild conversation with acceptable processing time and accuracy, where by acceptable is implied that the delay in processing users' inputs and respond to them does not prejudice the human experience in interacting with such systems. The motivation is to deploy the resulting application in classrooms as on of the future steps. To simplify this complex problem, we will start by using a laptop instead of a robot and adults instead of children. To increase the accuracy of the Audio Speech Recognition (ASR), we will use lip reading since with this algorithms it has been demonstrated that, in extreme cases where the sound can absolutely not be used, it can still achieve a Word Error Rate (WER) of less than 40%.

- Finalization of the development of the robot digital twin- Design automatic software tests in the simulator- Teaching two new interns how to use and improve the simulator

- Creation a digital twin of our latest generation robot in a simulation environment (CoppeliaSim)to facilitate continuous integration and testing- Assembly of new robots- Benchmarking of a thermal camera for anomaly detection- Technical support for client demos