Skip to main content
. 2012 Jul 23;6:6. doi: 10.3389/fnbot.2012.00006

Figure 8.

Figure 8

States and actions of the robotic finger during the reinforcement learning tasks. Left: 6 × 6 areas in normalized encoder-position space represent the discrete reinforcement learning states. One thousand continuous encoder values (gray dots) obtained from a random policy indicate the finger's movement range in the state space. Right: eight motor actions set the PDIP tension and MCP torque.