Haptic perception in human robotic systems

Heinz Wörn, Catherina R. Burghart, Karsten Weiß and Dirk Göger


Why is haptic perception essential in human robotic systems? This question is often posed in connection with humanoid robots. First of all, humanoid robots are intended to assist people in a typical human environment. A person expects a humanoid robot to think, move, act, and communicate in a human-like manner. This also includes the usage of typical human senses like vision, hearing and tactile sensing. Second, a robot manipulating items in an unstructured environment like a person’s home needs to have some haptic feedback: information whether an item is firmly grasped or sliding is important for handling objects. Third, different kinds of haptic and tactile feedback are required for moving and acting in a human-built environment: collision detection just as well as haptic feedback for actions or control by a human operator.
Different types of haptic feedback require an acquisition of environmental data by different kinds of haptic sensors. Most commonly, force torque sensors in the wrists or other joints of the robot are used. In addition, tactile input is achieved by sensor matrices based on various physical working principles. Sometimes, data inputs of depth perception and surface perception have to be combined to achieve an optimal haptic sensing.
This chapter mainly focuses on tactile sensing of a humanoid robot, illustrating the working principle of resistive sensor arrays used as an artificial sensitive skin, classifying different types of contact, and giving an insight into a current project of a humanoid robot.

Tactile sensor system

A tactile sensor system normally consists of discrete sensor cells, so called ‘texels’. They are arranged in homogeneous matrices, detecting an applied load profile. For data acquisition, the measurement converter is connected to a local intelligence, a sensor controller, digitising the sensor signals and pre-processing them. A host system processes the data made available by the controller and extracts characteristics. The data can be used, e.g., for reactive control of a robot. The measurement principles of the tactile sensor cells found in literature are based on three major classes: optical, capacitive and resistive effects. Optical sensors commonly utilise force dependent absorption or reflection of light beams [1]. They are very insensitive against corrosion and electromagnetic disturbances. For high area applications, like covering a whole robot system, the interconnection between the texels, commonly done by PMMA fibre cables, and the detection circuit becomes too complex. Another common approach for optical tactile sensors is to measure the scattered light from a lightened transparent polymer material by using a CCD camera [2]. These sensors can be used for tactile object recognition and orientation sensing in grippers, but are not suitable for surface covering.