Reaching in Developmental Robotics: Computational principles of infant reaching strategies in developmental robotics

Undergraduate Honors Thesis

Citations

Citeable URL: https://scholar.colorado.edu/concern/undergraduate_honors_theses/5h73px57h

Abstract

The field of developmental robotics stands to benefit from recent advances in the study of the development of reaching. These include verification of the minimum jerk hypothesis, the discovery that the adult nervous system finds excess degrees of freedom to be a luxury, clarifications as to the topographic organization of motor cortex, as well as a collection of datapoints that allow for the principled development and analysis of robotics models of infant reaching. In order to demonstrate the validity of these principles, a robot was developed using the Emergent Neural Network Simulator. The robot accounts for recent findings regarding motor cortex, solves the inverse kinematics of motion and learns using the biologically plausible Primary Value Learned Value reinforcement learning system, a first for the field. The flexibility of this framework allows for future, more realistic robots to be developed, including those that learn via autonomous mental development, the automatic selection of tasks based on intrinsic motivations.

Creator