A New Exploration Strategy for Soft Robots Based on Proprioception

Abstract

Exploring complex, unstructured environments requires a large set of information acquired by several sensors. Besides, a high level control is necessary to transfer and elaborate the whole data. Taking inspiration from plants, we aim at developing a system that is able to explore its surrounding environment paying a very low cost in terms of computation and data processing. Here, we demonstrate how a soft manipulator can identify the presence of an obstacle, or of a supporting structure, by means of a single sensor used together with a simplified mathematical model. The model can compute the configuration of the system only if no other forces other than the actuation are applied on it, while the sensor readings alone are not sufficient to infer the configuration. Exploiting the limitation of the model, an occurring mismatch between the expected position (computed by the model) and the measured one (by the sensor) provide sufficient information to identify the contact with a possible supporting structure. We validated the proposed methods in different scenarios in which we considered free motion without any obstacles, and in the presence of possible supports and impassable structures. To avoid false detection, we also considered the frequency of the discrepancy. In all the test cases the result is promising and this can pave the way toward a simplified approach for plant-inspired, non-vision-based navigation for soft devices.