Adaptive visual-motor coordination in multijoint robots using parallel architecture

摘要: This article derives and simulates a neural-like network architecture that adaptively controls visually guided, two-jointed robot arm to reach spot targets in three dimensions. The learns maintains visual-motor calibrations by itself, starting with only loosely defined relationships. geometry of the is composed distributed, interleaved combinations actuator inputs. It fault tolerant uses analog processing. Learning achieved modifying distributions input weights after each positioning. Modifications are made incrementally according errors consistency between signals used orient cameras those move arm. Computer simulations show intended acutator learning an average 4.3% signal range, across all possible targets.