EIT-NN: Data-driven EIT framework toward super-resolution of EIT-based robotic tactile sensings
Electrical-impedance-tomography (EIT)-based tactile sensor offers significant benefits on practical deployment because of its sparse electrode allocation, including durability, large-area scalability, and low fabrication cost, but the degradation of a tactile spatial resolution has remained challenging. A deep-neural-network-based EIT reconstruction framework, the EIT-neural network (EIT-NN), alleviates this tradeoff between tactile sensing performance and hardware simplicity. EIT-NN learns a computationally-efficient, nonlinear reconstruction attribute, achieving high-resolution tactile sensation and well-generalized reconstruction capability to address arbitrary complex touch modalities. EIT-NN consists of a uniquely suited network architecture design and novel loss function named “spatial-sensitivityaware mean-squared error (SSA-MSE)”. The network architecture adopted a two-fold transformation, decoding the sensor readings with fully-connected layers and image synthesis using CNN. The proposed SSA-MSE loss function uses knowledge about the spatial sensitivity of the sensor domain, guaranteeing a well-posed reconstruction with additional merits of smooth image generation and robust model training.
Related paper [2021 T-RO]
Adaptive and Optimal Measurement Algorithm for ERT-based Large-area Tactile Sensors
- The performace of ERT-based sensors is improved by increasing the number of electrodes, but the number of measurements and the conputational cost also increase.
- We propose an adaptive and optimal measurement algorithm for ERT-based tactile sensors.
- The measurement pattern consists of base pattern and local pattern.
- The tactile events are detected by a base pattern that maximizes the distinguishability of local conductivity changes.
- A set of local patterns are seletively recruited near the stimulated region to acquire more detailed information.
- The algorithm was implemented with a field-programmable gate array (FPGA).
Related paper [2021 T-Mech]
A large-area robotic skin for human robot tactile communication
- A robotic skin is developed using sparsely embedded microphones with soft materials, to easily conform into a large-curved surface, and to efficiently sense a touch using a few transducers.
- The touch is interpreted by classifying how the skin is touched (such as patting or slapping), and localising where the skin it touched.
- Such an interpretation would allow a tactile communication in between a human and a robot, as if human can socially interact with one another by interpreting a touch based on its essential information: how and where the touch is applied.
- For example, patting on one’s back is considered as an encouragement whereas slapping on back or patting on head is considered to be rude.
- Thus the developed skin may enrich the social interaction with any robotic system that physically interacts with a human.
Related paper [ICRA 2021]