Replicating Data Collection of Bayesian Optimization on a Six-Bar Tensegrity Robot

A soft robot, as opposed to a more conventional robot, is structurally unique as it is composed of flexible materials and its formation can be altered without damage. A specific type of robot that lies in the field of soft robotics is the tensegrity robot. It is a unique form of architecture which balances itself mechanically because of the way in which tensional and compressive forces are distributed and equilibrate with each other. The tensegrity robot at Union College is different from other tensegrities in that it is made with springs. It moves through vibration which is created by the spinning motors and oscillating springs. While this type of structure has advantages including the ability to maneuver into difficult to reach areas, portability, and cost, one of the main challenges with the tensegrity is movement controllability. One method proposed by previous researchers to circumvent this problem was to implement a Bayesian Optimization algorithm. Our goal for the summer was to run a variety of tests using Bayesian Optimization to determine whether this algorithm is optimal in finding gaits for a tensegrity’s movement. We collected a great amount of data, and were able to visually observe the robot moving a greater distance forward when the motors were set to high frequencies. We found this to be in contrast to when we set the motors to lower frequencies, resulting in almost no movement.