Controlling Execution Sequence Using Tactile-Classification During Manipulation by a Humanoid Robot

It is necessary to track system state during robotic manipulation. System state is defined to be manipulator and environment configuration. Without tracking, the robot is ignorant of the outcomes of its actions. State tracking enables the robot to respond to sub-task failures. Tracking system state is difficult for complex robots that incorporate hundreds of individual sensor signals. It is difficult to determine which signals are relevant and which are not, and to find low dimensional representations of system state.