Figure 2 | Scientific Reports

Figure 2

From: Intrinsic interactive reinforcement learning – Using error-related potentials for real world human-robot interaction

Figure 2

Scenario concept and task procedures. Each scenario contains a training phase to train a classifier and a test phase to evaluate this trained classifier. The reason for such classifier transfer is to reduce the calibration time. In the training phase, the subject observes the robot’s actions without interacting with the robot, i.e., without performing gestures (observation task). In this way, the time of data collection was substantially reduced in the training phase compared to the test phase that required an interaction with the robot (interaction task) by using gestures. In other words, we used the classifier trained on observation ErrPs to online detect interaction ErrPs in the test phase.

Back to article page