Robotics research has made strides to make robotic arms more maneuverable, lightweight, and cost effective. Fulfilling these objectives in simplest form is a soft robot consisting soft actuators. Soft actuators are lightweight flexible materials that are used for robotic applications. Since this category of robot has a simple design, the components available create limitations to the performance ability of the robot. This makes it challenging to fabricate a design that goes beyond standard procedures. This research examines a soft robot manipulator using twisted string actuators (TSAs). TSAs are a method used to create movements in robots by using a motor to twist and constrict a string to cause motion. TSAs are commonly used mechanics for creating movements but its application to soft robots are finite. This study explores the implementation of TSAs within a soft robotic arm. The soft robot arm limitations will be tested using the TSA to measure its maximum range of motion, bending angle, force of the soft gripper, and number of rotations. The arm gripper will perform on several objects and tip force sensors. Results show that the test model can produce a maximum tip force of about 500 mN and bending angle of 97.4°. Conclusions from this study may be used to further study soft robot mobility and its potential industrial application.

Multiple-input multiple-output (MIMO) wireless communication systems utilize numerous transmit and receive antennas to transmit multiple datapath signals simultaneously by taking advantage of the phenomenon known as multipath propagation, which results in multiple independent MIMO channels. Orthogonal frequency division multiplexing (OFDM) is a popular technique employed to encode data over these MIMO channels. As such, it is important to know the channel state information (CSI) for OFDM which is done by performing channel estimation to model some channel matrix. Here, we introduce an application of deep learning for channel estimation by describing an architecture for a variational autoencoder that is utilized to perform channel estimation on a 2x2 MIMO system and we compare it to other channel estimation techniques, such as least-squares estimation and minimum mean square error (MMSE) estimation. Additionally, a framework is introduced for developing a federated learning scheme for efficient channel estimation for MIMO systems.