A visual servo reinforcement learning control of uncalibrated manipulators with multi-channel gain decision | Synapse