Reinforcement Learning For Control Of A Multi-Input, Multi-Output Model Of The Human Arm