Curiosity-driven learning of locomotion. Driven by intrinsic motivation, the robot explores how movements of its legs make its whole body move around. Initially, he has no knowledge of its own body and of the environment (but his body has a bio-inspired morphology and motor primitives, which provide initial structure). He is not programmed to learn specific movements such as going forward or turning on the right or left, but rather chooses to experiment what he finds "interesting": he is self-motivated to explore movements which produce "learning progress", i.e. which allows it to improve its knowledge or its skills per se. At the beginning, movements are somewhat random. Then, they robot focuses on certain kinds of movements which produce initially high-learning progress, such as walking backwards. Then, it focuses on learning movements that make it walk a bit like a crab and turn. Finally, he explores and learn movements that make it crawl forward.
"
Intrinsic Motivation Systems for Autonomous Mental Development",
IEEE Transactions on Evolutionary Computation, 11(2), pp. 265--286.
by Oudeyer P-Y, Kaplan , F. and Hafner, V.
April 2007
"
Active learning of inverse models with intrinsically motivated goal exploration in robots"
Robotics and Autonomous Systems, 61(1), pp. 49-73
by Baranes, A., Oudeyer, P-Y.
September 14, 2011