This paper presents a theory and computer simulation of a neural controller that learns to accurately move and position a link carrying an unforeseen payload. The neural controller learns adaptive dynamic control from its own experience. It does not use information about link mass, link length, direction of gravity, and uses only indirect, uncalibrated information about payload and actuator limits. Its average positioning accuracy across a large range of payloads after learning is 3 percent of the positioning range. This neural controller can be used as a basis for coordinating any number of sensory inputs with limbs of any number of joints. The feedforward nature of control will allow parallel implementation in real time across multiple joints.