Pneumatic coding blocks enable programmability of electronics-free fluidic soft robots

Abstract

Decision-making based on environmental cues is a crucial feature of autonomous systems. Embodying this feature in soft robots poses nontrivial challenges on both hardware and software that can undermine the simplicity and autonomy of such devices. Existing pneumatic electronics-free soft robots have so far mostly been approached by using system fluidic circuit architectures analogous to digital electronics. Instead, here we design dedicated pneumatic coding blocks equivalent to If, If...break, and For software control statements, which are based on the analog nature of nonlinear mechanical components. We demonstrate that we can combine these coding blocks into programs to implement sequences and to control an electronics-free autonomous soft gripper that switches between behaviors based on interactions with the environment. As such, our strategy provides an alternative approach to designing complex behavior in soft robotics that is more reminiscent of how functionalities are also encoded in the body of living systems.

Fig. 1.. Overview of our design approach to enable pneumatic coding.

Our approach is based on a library of dedicated pneumatic circuits that represent specific coding statements. These statements can be combined to develop pneumatic programs that can respond to the environment.

Fig. 2.. Pneumatic equivalent of an If statement.

(A) Experimental realization of a soft hysteretic valve in its holder in its closed and open states. Scale bar, 12 mm. (B) Experimental realization of the If pneumatic circuit. (C) Measurement of the pressure $p$ upon interacting with (squeezing) the compliant volume. The interaction occurs at $t=0$ s. The green and black dashed lines indicate $p_{\text{in}}=35$ kPa and $\mathrm{\Delta }{p}_{\text{open}}=55$ kPa of the hysteretic valve, respectively. The gray solid lines are individual pressure measurements at 10 different interaction events, and the black solid line indicates their average. The inset photos represent the state of the soft actuator at the specific moment in time. (D) Schematic of pneumatic circuit used for the If functionality in (B).

Fig. 3.. Pneumatic equivalent of an If...break statement.

(A) Schematic of the components used to assemble a pneumatic N.O. valve. (B) Assembled N.O. valve, for different gate pressures. (C) Schematic of the If...break. (D) Experimental measurement for a pneumatic circuit based on the schematic in (C), for an interaction event with the compliant volume that occurs at $t=0$ s. The black, blue, and green solid lines represent the approximately constant input pressure $p_{\text{in}}\approx 68$ kPa, the pressure at the gate $p_{\mathrm{G}}$ of the N.O. valve, and the actuator pressure $p$, respectively. The dashed line represents the opening pressure $\mathrm{\Delta }{p}_{\text{open}}=70$ kPa of the hysteretic valve. Snapshots of the actuator’s state are taken from the experimental realization of the setup at the corresponding moments in time.

Fig. 4.. Pneumatic equivalent of a For statement.

(A) Schematic of the pneumatic For circuit. $x$ indicates the linear extension of the soft actuator. (B) Experimental measurement of the averaged pressure in volumes $V_1$ and $V_2$ and tracking of the extension $\mathrm{\Delta }x$ of the linear actuator with respect to its resting position over 20 different repetitions of the For loop. (C) Snapshots of the linear actuator used for the corresponding highlighted positions as indicated by the colored dots in (B). (D) Different number of iterations (left) and periods (right) of the For loop for different volume combinations of $V_1$ and $V_2$. The black entries on the heatmaps represent $V_1$ and $V_2$ combinations for which the For behavior in (B) was not observed. Note that, for this experiment, the soft actuator was removed from the setup. (E) Schematic and (F) experimental data of the nested For loop.

Fig. 5.. Schematics of a pneumatic circuit and component to obtain sequential execution of instructions.

The For loop enables clocking capabilities and feeds a 0.1 mm PVC punched card that encodes the instructions to be executed. By connecting an input pressure $p_{\text{in}}=367$ kPa, the four different output channels can be pressurized following the sequence of instructions in the physical support memory. (A) Circuit design and physical support memory. In black, the punched holes provide the change of fluidic resistance required to pressurize a specific channel. The duration of operation can be encoded in the different lengths of each aperture. (B) Experimental realization of the setup. (C) Highlight of bellow deformation as a consequence of the sequential readout of instructions in time. We report the data from this experiment in fig. S4 and its realization in movie S1.

Fig. 6.. Encoding pneumatic continuous variables using N.O. valves.

(A) Schematic of the pneumatic circuit. (B) Isofrequency $f_{\text{Act}}$ and (C) isoamplitude of oscillations $\mathrm{\Delta }{p}_{\text{Act}}$ curves as a function of the control parameters $p_{\mathrm{G},\text{Ampl}.}$ and $p_{\mathrm{G},\mathrm{f}}$.

Fig. 7.. Assembly of pneumatic coding blocks to design a circuit for a soft robotic retriever demonstrator.

(A) Highlight of the hardware components. We assemble a soft linear actuator with two bending actuators through a PLA frame. This allows us to obtain a pneumatic soft robot that can elongate and grab target objects. (B) We combine the For (blue) and If...break (red) pneumatic coding blocks to functionalize the actuators so that the pneumatic soft retriever executes the program in Algorithm 6.

Fig. 8.. Experimental data and schematic representation of the obtained behavior of the soft robotic retriever demonstrator.

(A to C) Pneumatic responses of the circuit before and after the If condition is triggered upon gripping the orange object for the bending actuator branch (green), the linear actuator branch (blue), and the feedback branch (red), respectively. In the top row, we report an overview of the two different pneumatic responses before and after the If...break condition is triggered, with the successful gripping event occurring at $t^{*}=368$ s. The shaded time window of the interaction event is shown in more detail in the bottom row. (D to G) Schematic representation of the behavior of the pneumatic soft retriever unit running Algorithm 6. Highlights of behavior before and after the If...break condition is triggered are highlighted respectively in red and green.