Learning path: technology, artificial intelligence/machine learning/developmental robotics
This path is usually characterised by these features:
- Embodied computational models: this means that the model is tested in an embodied agent, i.e. a simulated or real robot that interacts with the world through its sensors and actuators (the core of embodiment is that the interaction is circular: the actions change the environment and this affects the next sensation).
- The overall goal of the research is usually technological: here one aims to buld a better robot; usually within LOCEN the controller we work on are strongly inspired by data from developmental psychology, but the goal is still technological. This for example implies that the systems created have to be more powerful that those existing in the robotic literature.
- The overall goal is sometimes technological (typical of developmental robotics): here one selects a one or more experimental data, usually behavioural, from developmental psychology, and tries to reproduce them with the model. In so doing , the model produces a possible interpretation of the data, and possibly suggests some further experiments as it produces some predictions to test.
The key pieces of knowledge to acquire are these:
- Developmental robotics. Ask some key papers from LOCEN members on:
- Methodology of developmental robotics
- Specific papers on the area and problem of investigation you selected
The skills to learn are these:
Decide with the PI and the senior LOCEN researcher which tools to use for programming and for implementing the simulated/real robot:
- Programming in Python: expecially in the case the objective is to target empirical data; however, we tend more and more to use this language even in the case of technological goals
- Programing in C++: you have to learn this if the speed of simulation is critical
- Physical simulator: this is needed to work with a simulated robot to test the model
- Real robot: in case you will have to use a real robot