Reprogrammable sequencing for physically intelligent under-actuated robots

TL;DR

The paper addresses navigation in unstructured environments with minimal electronics by embedding physical intelligence into reprogrammable under-actuated units that sequence motions via multistable energy landscapes. The core method defines a unit cell with $\mathcal{E}(\theta,\boldsymbol{p},\boldsymbol{q})=\frac{k}{2}[\ell(\theta,\boldsymbol{p},\boldsymbol{q})-\ell_0]^2$ and programmable anchor structure, then demonstrates how serial coupling and environmental interactions reprogram the locomotion sequence under a single quasi-static actuator. A four-DOF robot demonstrates forward, backward, and turning gaits, tunable by anchor points and environmental input; an antenna-based sensing module enables gait adaptation in response to mechanical contacts without electronic sensors, and scalability to more DOFs and monolithic fabrication is discussed. The work offers a pathway to physically intelligent robots with reduced sensing/actuation hardware and potential for rapid, snapping-based motions, expanding the scope of passive mechanical computation in robotics.

Abstract

Programming physical intelligence into mechanisms holds great promise for machines that can accomplish tasks such as navigation of unstructured environments while utilizing a minimal amount of computational resources and electronic components. In this study, we introduce a novel design approach for physically intelligent under-actuated mechanisms capable of autonomously adjusting their motion in response to environmental interactions. Specifically, multistability is harnessed to sequence the motion of different degrees of freedom in a programmed order. A key aspect of this approach is that these sequences can be passively reprogrammed through mechanical stimuli that arise from interactions with the environment. To showcase our approach, we construct a four degree of freedom robot capable of autonomously navigating mazes and moving away from obstacles. Remarkably, this robot operates without relying on traditional computational architectures and utilizes only a single linear actuator.

Figures (4)

• Figure 1: (A) Schematic representation of a unit cell. (B) Picture of a unit cell in its two possible stable states, state 0 and state 1. (C) Schematic of a unit cell with six different rubber band configurations, all defined by $\mathbf{p}$ = [-1, 0] cm, but varying $\mathbf{q}$ = [10,0], [0,0], [20,0], [10,-15], [20,15] and [0,-15] mm (black, red,blue, yellow, green, purple, respectively). (D)-(E) Evolution of the vertical (D) reaction force $F$ and (E) strain energy $\Delta\mathcal{E}$ as a function of the rotation of the levers $\theta$. The colors correspond to the configurations shown in (C).
• Figure 2: (A) Schematic of a mechanism comprised of two unit cells connected in series. (B) The four possible states and diagram highlighting the four observed transition sequences. (C) Effect of $\mathbf{q}^{out}$ on the sequence of transitions for the mechanism in (A) with $\mathbf{p}^{in} = −\mathbf{q}^{in} =$ [−10, 0] mm. The colors correspond to the sequences highlighted in (B) and the light shaded areas denote unit cells that are monostable when considered individually. The markers highlight mechanisms built choosing from the five different configurations of the unit cell shown in Fig. \ref{['fig:single unit_main']}C as the outer unit. (D)-(G) We consider four of the two-unit mechanisms indicated by a marker in (C) and for each of them show a schematic (top), the energy landscape together with the supported state transitions (center), and the force-displacement response measured in experiments and predicted by theory.
• Figure 3: (A) Picture and schematic of the under-actuated robot. (B) Top and side view of one leg of the robot as it moves through its four supported states. Moving the inner unit from state 0 to state 1 propels the entire mechanism forward, whereas shifting the outer unit from state 0 to state 1 raises the foot. (C) Trajectory followed by the robot for $\mathbf{q}^{out} = [20,0]$ mm (left) and $\mathbf{q}^{out} = [5,0]$ mm (center) on both sides, and $\mathbf{q}^{out} =$[20,0] mm and $\mathbf{q}^{out} =$[5,0] mm on the left and right side, respectively (right). The initial position corresponds to the white dashed outlines, while the picture indicate the position of the robot after $7\frac{3}{4}$ cycles. (D)-(E) Evolution of the (D) displacement, $d$, and rotation, $\varphi$, of the robot for the three configurations of the rubber bands considered in (C).
• Figure 4: (A) Picture of the robot with a module with two antennas attached at its front. (B) Schematic of showing that the movement of the right antenna alters $\mathbf{p}^{in}$ of the left mechanism. (C) Snapshots of the robot interacting with a cylindrical obstacle (see also Video S4) (D) Trajectory of the robot navigating through an environment obstructed by three walls.