Artificial skin integrating an innovative tactile sensor network on a grape harvesting robotic hand

Gentle handling of objects requires a sense of touch, which has not been completely solved by the robotic world. This thesis proposes a novel tactile sensor network, acting as an artificial skin, specifically tailored for grape harvesting tasks. A sensing mechanism, based on two different operating principles, magnetic and optical, offers an extremely sensitive solution needed to accomplish the grabbing maneuver without damaging the fruit. Its complex shape makes evident the necessity of a highly scalable framework, whose number of sensors could increase rapidly, such that as much surface as possible is covered. An innovative approach to serial communication is presented in this study. The whole system works without using common multi-point protocols like SPI or I2C. A clock, enabling a flip-flop, and an input signal, giving the synchronization, are the core of the scheme. The nodes, connected in a daisy chain, are empowered one at a time, allowing a straightforward data acquisition procedure. An entirely operative module composed of four sensing elements has been developed. Three variants have been built sharing the same circuit idea: an IC with four FFs mainly constitutes the bottom side of the PCB. Their difference lies in the sensing components. The fully magnetic one is characterized by four Hall Effect sensors soldered on top and by a silicone layer with four perfectly aligned magnets. When a force compresses the rubber silicone, the magnets in correspondence to the contact point get closer to the sensors below. The optical module instead exploits four optocouplers: infrared light generated by a LED is reflected by the silicone surface and measured by a phototransistor, placed next to the LED in the same chip. Both techniques transduce pressure into distance, which is detected in the former case by a magnetic field variation, while in the latter one by a light absorption difference. The last version is a hybrid solution, which encloses two Hall effect sensors and two optocouplers. Its aim was to combine both sensitivity capabilities. In all configurations, the network interacts with a custom board, whose MCU manages data collection and transmissions through BLE. This work explains in detail the firmware used to drive the system. Considering a total number of 199 sensing elements, a maximum reading frequency of 10.05Hz per sensor has been achieved. This performance in terms of real-time monitoring has been successfully tested on the final network, based merely on Hall effect sensors for their reliability. Once the hand shape was chosen, ad-hoc flexible PCBs were designed in order to make even curved areas, such as fingertips, responsive to touch. With the purpose of simplifying the oversight, a data visualization interface, representing each sensor in the system according to its position, has been created. Successful tests of the final skin prototype, which softly grasps grapes while perceiving contact, lend credence to the project.