ALL WE CAN SAY ABOUT ROBOTICS
What is paradoxical is that robotics is producing unprecedented programmable objects, but quite different from the usual fantasy of science-fiction of the last century. It plunges the digital world into the very heart of the real world, and beyond the accepted ideas ..
- Algorithm of learning: this corresponds to algorithms that adjust the parameters of their calculations according to the examples given to them, or returns (positive or negative, like rewards or punishments) resulting from previous calculations. This makes it possible to adapt their operation to the data provided.
- Algorithm of optimization: it corresponds to algorithms which solve a problem by successive improvements: one starts from an initial solution by default, one modifies it a little in one direction or another, and if one of its modifications improves the solution the process is repeated. A gain criterion to be maximized or a cost to be minimized is therefore the basis of these methods.
- Optimization by trial / error: this corresponds to algorithms that adjust the parameters by « reinforcement », a solution is tested and depending on the positive or negative feedback, an adjustment is made to move towards a better behavior. The fact that the return happens after the action that may have been the cause, sometimes much later, requires having an internal representation of what happens over time.
- Modeling: representation of a real or abstract object, eliminating the difficult or accessory details to reproduce, in order to obtain a clearer result to interpret, a model is validated by its adequacy to data (different from those which could help to build it), thus to predict new facts; it may be mathematical or computer, but also be a tangible object (a model, a model animal for some biological functions).
- Brain: it is the main organ of the nervous system of the animals, it regulates the other systems of organs of the body, acting on the muscles or the glands, and constitutes the seat of the cognitive functions (memory, learning, planning, etc.). ); this cognition exists only because the brain is embodied in a body interacting with an environment.
ABOUT THE AUTOR
Pierre-Yves Oudeyer is Inria Research Director and Scientific Leader of the FLOWERS team. A multi-award winning Artificial Intelligence Doctor, he is interested in computer and robotics modeling of sensorimotor and social learning and development in humans and machines. In particular, he studies the mechanisms of curiosity, maturation, the role of the body in cognitive development, and human-robot interactions that allow learning by imitation.