Soft robots have been extensively researched due to their flexible, deformable, and adaptive characteristics. However, compared to rigid robots, soft robots have issues in modeling, calibration, and control in that the innate characteristics of the soft materials can cause complex behaviors due to non-linearity and hysteresis. To overcome these limitations, recent studies have applied various approaches based on machine learning. This paper presents existing machine learning techniques in the soft robotic fields and categorizes the implementation of machine learning approaches in different soft robotic applications, which include soft sensors, soft actuators, and applications such as soft wearable robots. An analysis of the trends of different machine learning approaches with respect to different types of soft robot applications is presented; in addition to the current limitations in the research field, followed by a summary of the existing machine learning methods for soft robots.

Soft sensors made of highly deformable materials are one of the enabling technologies to various soft robotic systems, such as soft mobile robots, soft wearable robots, and soft grippers. However, major drawbacks of soft sensors compared with traditional sensors are their nonlinearity and hysteresis in response, which are common especially in microfluidic soft sensors. In this research, we propose to address the above issues of soft sensors by taking advantage of deep learning. We implemented a hierarchical recurrent sensing network, a type of recurrent neural network model, to the calibration of soft sensors for estimating the magnitude and the location of a contact pressure simultaneously. The proposed approach in this paper were not only able to model the nonlinear characteristic with hysteresis of the pressure response, but also find the location of the pressure.

This paper describes a hybrid brain-computer interface (BCI) technique that combines the P300 potential, the steady state visually evoked potential (SSVEP), and event related de-synchronization (ERD) to solve a complicated multi-task problem consisting of humanoid robot navigation and control along with object recognition using a low-cost BCI system. Our approach enables subjects to control the navigation and exploration of a humanoid robot and recognize a desired object among candidates. This study aims to demonstrate the possibility of a hybrid BCI based on a low-cost system for a realistic and complex task. It also shows that the use of a simple image processing technique, combined with BCI, can further aid in making these complex tasks simpler. An experimental scenario is proposed in which a subject remotely controls a humanoid robot in a properly sized maze. The subject sees what the surrogate robot sees through visual feedback and can navigate the surrogate robot. While navigating, the robot encounters objects located in the maze. It then recognizes if the encountered object is of interest to the subject. The subject communicates with the robot through SSVEP and ERD-based BCIs to navigate and explore with the robot, and P300-based BCI to allow the surrogate robot recognize their favorites. Using several evaluation metrics, the performances of five subjects navigating the robot were quite comparable to manual keyboard control. During object recognition mode, favorite objects were successfully selected from two to four choices. Subjects conducted humanoid navigation and recognition tasks as if they embodied the robot. Analysis of the data supports the potential usefulness of the proposed hybrid BCI system for extended applications. This work presents an important implication for the future work that a hybridization of simple BCI protocols provide extended controllability to carry out complicated tasks even with a low-cost system.

Soft sensors are becoming more popular in wearables as a means of tracking human body motions due to their high stretchability and easy wearability. However, previous research not only was limited to only certain body parts but also showed problems in both calibration and processing of the sensor signals, which are caused by the high nonlinearity and hysteresis of the soft materials and also by misplacement and displacement of the sensors during motion. Although this problem can be alleviated through redundancy by employing an increased number of sensors, it will lay another burden of heavy processing and power consumption. Moreover, complete full-body motion tracking has not been achieved yet. Therefore, we propose use of deep learning for full-body motion sensing, which significantly increases efficiency in calibration of the soft sensor and estimation of the body motions. The sensing suit is made of stretchable fabric and contains 20 soft strain sensors distributed on both the upper and the lower extremities. Three athletic motions were tested with a human subject, and the proposed learning-based calibration and mapping method showed a higher accuracy than traditional methods that are mainly based on mathematical estimation, such as linear regression.

Visual information plays an essential role in enhancing neural activity during mental practices. Previous research has shown that using different visual scenarios during mental practices that involve imagining the movement of a specific body part may result in differences in performance. Many of these scenarios utilize the concept of embodiment, or one’s observation of another entity to be a part of oneself, to improve practice quality of the imagined body movement. We therefore hypothesized that applying immersive virtual reality headsets, with their ability to provide rich immersion and illusion by presenting egocentrically simulated virtual scenarios, and action observation to motor imagery practice will result in significant improvement. To explore the possible synergy between immersive systems and motor imagery, we analyzed the electroencephalogram signals of our participants as they were presented the same virtual hand movement scenario with two different mediums: an immersive virtual reality headset and a monitor display. Our experimental results provide evidence that the immersive virtual reality headsets induced improved rhythmic patterns with better discriminating spatial features from the brain compared to the monitor display. These findings suggest that the use of immersive virtual reality headsets, with the illusion and embodiment they provide, can effectively improve motor imagery training.