Detecting scalable obstacles using soft sensors in the body of a compliant quadruped

Abstract

In soft robotics, one of the trending topics is using soft sensors to have feedback from the robot's body. This is not an easy process to accomplish since the sensors are often nonlinear, so researchers use different methods to generate information from data such as filters, machine learning algorithms, and optimization algorithms. In this paper, we show that, with good electronic and mechanical design, it is possible to use soft sensors for detecting obstacles and distinguishing the scalable obstacles. The demonstration is conducted with an untethered miniature, soft, C-legged robot, M–SQuad, the first modular C-legged quadruped consisting of three modules, which are connected by four soft sensors. In M–SQuad's body design, sensors are utilized as both sensing and structural elements. The modular design of the M–SQuad allows testing different sensor geometries and replacing the malfunctioning parts easily, without the need to refabricate the entire robot. A case study is introduced for demonstration of the robot's capability of detecting obstacles and distinguishing scalable obstacles in a parkour consisting of two obstacles with the heights of 20 mm and 150 mm, respectively. In the case study, M–SQuad can detect an obstacle during locomotion using the coil-spring shaped soft sensors in its body. Moreover, it can distinguish the obstacle is scalable or not after an initial climbing trial. If the obstacle is not scalable, the robot turns back.