Origami-based integration of robots that sense, decide, and respond

The paper addresses the challenge of achieving full sense-decide-act autonomy in origami robots without relying on rigid, semiconductor-based electronics. Traditional origami robots often require external electronics for sensing, control, and actuation, which adds weight, complexity, and integration challenges due to stiffness mismatch and vulnerability to harsh environments. The authors posit that embedding sensing, computing, and actuation directly into compliant, conductive materials could enable autonomous origami robots that retain origami’s advantages (rapid prototyping, low cost, accessibility) while operating robustly. They aim to develop a compliant-material-based architecture that processes information and interfaces seamlessly with origami sensors and actuators to realize complete autonomous loops.

The authors review efforts integrating smart materials into origami for sensing, computing, communication, and actuation, as part of a broader push toward non-traditional computing (mechanical, pneumatic, fluidic, and metamaterial-based logic) in soft robotics, microfluidics, and mechanics. Prior work demonstrated individual components (e.g., foldable mechanical logic, soft ring oscillators, fluidic logic, metamaterial logic), but integrated autonomous origami systems remain elusive due to a lack of suitable computing elements that can interface with compliant sensors and actuators, high resistance/energy loss in current components, and fabrication complexity. The paper situates its contribution in delivering a low-resistance, cascadable, compliant computing element that enables seamless integration with sensing and actuation.

The core enabling mechanism is the Origami Multiplexed Switch (OMS), a compliant 2-to-1 multiplexer built from a bistable beam and two conductive super-coiled polymer (CSCP) thermal actuators. Each actuator, driven by Joule heating, pulls the bistable beam to snap between two stable states, toggling electrical contacts (tabs/poles) and selecting between two inputs VR and VS to produce output Q according to Q = VR·V1 + VS·V2. The beam’s snap-through provides binary states and hysteresis, enabling power usage only during switching and robustness to signal noise. Tabs on the bistable beam are optimized (width 1.2 mm, length 4.0 mm, folded angle −30°) to ensure reliable contact, yielding ON resistance ~0.6 Ω and OFF resistance ~1.2 MΩ (ON/OFF ~10^6), supporting low-voltage actuation (<3 V) and large fan-out. Characterization measured gate delay versus selection voltage and cooling conditions: typical operation at VS=2.4 V shows ~1.5 s gate delay; threshold ~1.19 V in still air (lower bound delay ~0.1 s) and ~1.8 V with active cooling; simplified thermal models provided fitted thermal mass and conductivity. The OMS also functions as a relay, controlling outputs at higher supply voltages than the control signal, and exhibits noise immunity due to hysteresis. Building on OMS, the authors implement origami digital logic (NOT, AND, OR) and combinational gates (NAND, NOR) by reconfiguring input assignments and cascading OMS units. A nonvolatile origami set-reset latch provides 1-bit memory with write/erase and holds state through power outages; hold time is ~1.5 s. Three integrated robots demonstrate sense-decide-act loops: (1) a Venus flytrap-inspired robot using two origami touch sensors (bistable-beam derived) feeding an AND gate to drive CSCP actuators to close leaves only when both sensors are triggered; (2) an untethered crawler using tactile antenna sensors and an OMS-derived OR gate to actuate a CSCP-driven DPDT switch (modified OMS) to reverse DC motor polarity for collision avoidance; (3) a wheeled car reading reprogrammable OMS-based memory bits on a rotating disc via a read head to set wheel speeds and follow specified trajectories. Fabrication: Origami bodies from patterned 0.127 mm polyester film cut by plotter/laser; copper tape traces (0.05 mm) as circuits; CSCP actuators prepared by twisting conductive yarn, annealing (0.45 A, cyclic for 8 h), and training (0.27 A). Devices powered by DC supply; signals recorded via NI myDAQ; optional active cooling (~110 cfm, ~130 mm distance) used to reduce actuator cooling time.

- OMS achieves low ON resistance (~0.6 Ω) and high OFF resistance (~1.2 MΩ), giving ON/OFF ratio ~10^6 and enabling low-voltage control (<3 V) with large fan-out. - Typical gate delay ~1.5 s at VS=2.4 V; modeled lower bound delay ~0.1 s at high supply; threshold control voltage ~1.19 V (still air) and ~1.8 V (with active cooling). Fitted actuator thermal parameters: C_th≈4.48×10^-2 Ws/°C and λ≈1.13×10^-2 W/°C (still air); C_th≈4.62×10^-2 Ws/°C and λ≈2.58×10^-2 W/°C (cooling). - OMS functions as a relay: 2.4 V control pulses can switch outputs with supply up to 9.6 V (≈4× control), benefiting from input-control circuit isolation. - Noise robustness demonstrated: moderate noise on control does not alter binary output due to hysteresis. - Digital logic realized in compliant materials: NOT, AND, OR gates; cascaded NAND and NOR gates operating with full truth tables; XOR/XNOR designs indicated. - Nonvolatile origami S-R latch stores 1 bit with ~1.5 s hold time; retains state across power outages; supports write/erase/rewrite. - Flytrap-inspired robot autonomously captures “living prey” when both leaf-embedded touch sensors are triggered; observed actuation delay ~1.7 s and leaf closure in several seconds at 2.4 V. - Robust operation under adversarial conditions where semiconductor systems failed: static magnetic field 0.47 T, RF interference 5 W, electrostatic discharge ≥20 kV, and mechanical deformation up to 50°. - Untethered legged robot autonomously reverses upon obstacle detection using tactile sensors, OMS-based OR logic, and a CSCP-driven DPDT switch. - Wheeled car follows reprogrammable trajectories via OMS-based memory disc; with four 2-bit words (8 bits total), supports 256 trajectories; demonstrated letters “u”, “c”, “T”, and “a”. - Estimated maximum switching speed ~3.2 s considering ~3 s cooling time plus minimal gate delay (~0.1 s).

Embedding OMS-based sensing, logic, memory, and actuation into compliant, conductive materials addresses the core challenge of achieving autonomous behavior in origami robots without rigid electronics. The OMS provides a low-resistance, cascadable switching element that interfaces naturally with electrically mediated origami sensors and CSCP actuators, enabling robust digital logic, relays, and memory. Demonstrations of a flytrap-inspired gripper, a collision-avoiding crawler, and a memory-programmed car validate complete sense-decide-act loops on origami platforms, operating robustly under environmental conditions that degrade semiconductor electronics. These findings show that complex computation and control can be realized mechanically/electromechanically within folded structures, opening a pathway toward more sophisticated autonomous machines (e.g., finite state machines) while preserving origami’s advantages in cost, weight, manufacturability, and resilience.

The work introduces an integrated, semiconductor-free approach to autonomous origami robots through the Origami Multiplexed Switch and related architectures for computation and memory. It demonstrates compliant, low-loss, noise-robust logic and storage elements that can be cascaded, and integrates them with origami sensors and CSCP actuators into functional robots capable of autonomous interaction with their environment. Future directions include: increasing computation complexity (finite state machines and potentially Turing-complete architectures), improving speed via faster bistable mechanisms and actuator designs, automating memory writing (e.g., via CSCP actuation or magnetic control), expanding memory density and depth for richer trajectories and behaviors, integrating additional sensing modalities, and further reducing reliance on non-origami components (e.g., replacing DC motors for read heads with origami-compatible actuation).

- Switching speed is limited by thermal actuator heating/cooling and bistable snap dynamics (typical delays ~1.5 s at 2.4 V; maximum switching period ~3.2 s considering cooling), which is slower than semiconductor electronics. - Gates require explicit reset and actuator cooldown for continuous operation. - The flytrap robot does not process temporal information (e.g., two touches within a time window) as in the biological analog; capture speed is limited at low supply voltages. - Some demonstrations use conventional components (DC motors for locomotion and memory read head), so systems are not fully origami-only. - Memory writing on the car’s disc is manual in the current prototype; automated, programmable writing is proposed but not implemented. - Radiation robustness is inferred from known semiconductor vulnerabilities; direct radiation testing was not reported.