Electropneumatic Oscillators Using Nonlinear Inflatables

Animals and robots employ central pattern generators, networks that invoke rhythmic patterns from constant inputs, to orchestrate limb movements during locomotion. Artificial central pattern generators (CPGs) can be either implemented in software or constructed in a physical domain. The former lacks embodiment, inhibiting direct interactions with the physical world. The latter is restricted by a complex translation of abstract functions to the physical domain, e.g., negative feedback, if-then behavior, etc. Here, self-oscillators, a rudimentary type of central pattern generators, that find ground in both the pneumatic and electrical domains are demonstrated. First, the hysteretic behavior of previously developed conical membranes is analyzed in deformation space. This state-space behavior is then transformed to the electrical domain by means of a stretchable strain sensor. Next, an analog comparator distinguishes between the two states and instructs a pneumatic solenoid valve to counteract the current state. As a result, a stable oscillation emerges, with a frequency that is dominated by the physical characteristic of the pneumatic circuit. As such, the proposed electropneumatic oscillator provides a promising platform for building complex CPGs that control interactive neuromorphic robots.