Leo Pape

Dalle Molle Institute for Artificial Intelligence Research

Papers

Total Citations

H-Index

About

Leo Pape is a leading researcher in developmental robotics and autonomous skill acquisition, whose work bridges the gap between curiosity-driven learning and high-dimensional sensory processing. His most influential contribution, the 2012 paper "Learning tactile skills through curious exploration" (52 citations), pioneered a groundbreaking approach where a biomimetic robot finger, equipped with microelectromechanical touch sensors, autonomously acquires tactile exploratory skills through intrinsic motivation rather than task-specific programming. This curiosity-driven paradigm has become foundational in the field. Pape further advanced humanoid motion planning with his "Task-relevant roadmaps" framework (20 citations), introducing Natural Gradient Inverse Kinematics—a sampling-based optimizer leveraging natural evolution strategies—to enable complex, multi-degree-of-freedom robot motions. His work on sensory abstraction is equally notable: through Curiosity-Driven Modular Incremental Slow Feature Analysis (14 citations) and AutoIncSFA (8 citations), he developed systems that autonomously learn compact, meaningful representations from high-dimensional video streams, allowing humanoid robots to recognize spatio-temporal patterns like a person approaching. Pape’s research fundamentally redefines how robots can learn from their environment, moving from rigid programming to flexible, self-directed exploration.