Adding the context dependent parameters such as locations of start and end-points to the motor goal, iCub begins to mentally simulate the virtual trajectory for imitating the demonstrated trajectory by reusing the previously learnt parameters about the bump primitive. Hence reuse of past motor experience (in a different task) allows the robot to spontaneously imitate the teacher with a fair enough first prototype in a new situation. Some minor smoothening of the trajectory is needed around the already known parameters. Figure 5 Panel E – M show nine mentally simulated trajectories by iCub while fine tuning the parameters. Figure 5 Panel M shows the final solution reached which is re-enacted by the robot as shown in Panels N-P. A video recording of the described experiment on iCub is available at the following link. In sum, a teacher’s demonstration reduces the timeframe for exploration, past motor experience stored in the form of shape primitives provides the necessary parameters to instantaneously generate first prototype, which is further improvised by small modifications in the parameters around the already estimated
Adding the context dependent parameters such as locations of start and end-points to the motor goal, iCub begins to mentally simulate the virtual trajectory for imitating the demonstrated trajectory by reusing the previously learnt parameters about the bump primitive. Hence reuse of past motor experience (in a different task) allows the robot to spontaneously imitate the teacher with a fair enough first prototype in a new situation. Some minor smoothening of the trajectory is needed around the already known parameters. Figure 5 Panel E – M show nine mentally simulated trajectories by iCub while fine tuning the parameters. Figure 5 Panel M shows the final solution reached which is re-enacted by the robot as shown in Panels N-P. A video recording of the described experiment on iCub is available at the following link. In sum, a teacher’s demonstration reduces the timeframe for exploration, past motor experience stored in the form of shape primitives provides the necessary parameters to instantaneously generate first prototype, which is further improvised by small modifications in the parameters around the already estimated