Magnetic field-enhanced vertical integration enables embodied intelligence in untethered soft robots

Abstract

Embodied intelligence in soft robotics offers unprecedented capabilities for operating in uncertain, confined, and fragile environments that challenge conventional technologies. However, achieving true embodied intelligence-which requires continuous environmental sensing, real-time control, and autonomous decision-making-faces challenges in energy management and system integration. We developed deformation-resilient flexible batteries with enhanced performance under magnetic fields inherently present in magnetically actuated soft robots, with capacity retention after 200 cycles improved from 31.3 to 57.3%. These compliant batteries enable large-area deployment of 44.9% across the robot body, and their vertical integration with rationally designed flexible hybrid circuits minimizes additional stiffness while maintaining deformability. This actuator-battery-sensor vertical integration methodology maximizes functional area utilization in a manta ray-inspired soft robot, establishing an untethered platform with sensing, communication, and stable power supply. The system demonstrates embodied intelligence in aquatic environments through diverse capabilities including perturbation correction, obstacle avoidance, and temperature monitoring, with proprioceptive and environmental sensing enabling real-time decision-making during magnetically actuated locomotion.

Fig. 1.. Magnetic field–enhanced vertical integration enables embodied intelligence in untethered soft robots.

(A) Comparison between biological embodied intelligence, achieved through vertical integration of receptors, fat cells, and muscles, and conventional soft robots’ lateral integration of functional modules (sensory, actuation, and energy units). (B) Schematic of the manta ray–inspired soft robot demonstrating vertical integration of magnetic elastomer actuators, flexible Zn-MnO₂ batteries, and flexible hybrid circuits for synergistic locomotion, energy supply, and sensing. (C) Mechanism of magneto-electrochemical enhancement: the magnetic field from ferromagnetic elastomer suppresses both Zn dendrite growth and MnO₂ cathode deterioration in onboard battery. (D) Comparative capacity retention of flexible Zn-MnO₂ batteries after 200 cycles with and without magnetic field. The error bar indicates the SDs from the data of five independent samples. (E) Comparison of wing tip deflection and deformation stiffness (K_s) under applied manipulation magnetic fields across integration stages: (1) unloaded robot, (2) with flexible batteries, (3) with unoptimized circuit board, and (4) with optimized flexible hybrid circuit. The error bar indicates the SDs from the data of five independent samples. (F) Demonstration of embodied intelligence through digital twin–enabled obstacle avoidance and perturbation correction.

Fig. 2.. Magneto-electrochemical enhancement mechanism and characterization in Zn||Zn symmetric coin cells and Zn-MnO₂ coin cells.

(A) Illustrative comparison of Zn dendrite growth patterns with and without applied magnetic field. (B) GCD profiles of Zn||Zn symmetric coin cells at 2 mA cm⁻² under magnetic field influence. (C) Comparative analysis of life span and voltage hysteresis in Zn||Zn symmetric coin cells across different current densities, with and without magnetic field. The error bar indicates the SDs from the data of three samples. (D) SEM visualization of Zn electrode surface morphology after 300 cycles (scale bar, 10 μm). (E) Comparison of the average capacity of Zn-MnO₂ coin cells cycled at 0.2, 0.5, 1.0, 2.0, and 5.0 C across three test periods with and without magnetic field. The error bar indicates the SDs from the capacity data of five cycles. (F) GITT analysis after rate performance testing, comparing diffusion kinetics with and without magnetic field. (G and H) HRTEM lattice fringe images of MnO₂ cathode structure after 1000 cycles without (G) and with (H) magnetic field (scale bar, 1 nm). (I) Extended cycling performance comparison of Zn-MnO₂ coin cells under magnetic field influence. The thick solid lines and shaded bands indicate the means and SDs from the data of 5 samples. (J) Projected density of states (PDOS) analysis showing Mn 2p and O 2p orbital overlap in MnO₂ lattice under a magnetic field.

Fig. 3.. Magnetic field–enhanced vertical integration for untethered soft robot design.

(A) Orthogonal multiaxis magnetization of the manta ray soft robot: axial magnetization for steering and cosinusoidal magnetization for bending deformation. (B) CV analysis of flexible Zn-MnO₂ batteries under continuous vibration, comparing long-term cycling stability with and without magnetic field enhancement. (C) The average wing tip deflections of soft robots under a 15-mT magnetic field: unloaded (a₁ + a₂)/2 = 3.5 mm versus fully loaded (b₁ + b₂)/2 = 2.0 mm, 5 samples each. Scale bar, 1 cm. (D) Force-deflection curves of manta ray soft robots before and after vertical integration of flexible Zn-MnO₂ batteries and optimized flexible hybrid circuit. The thick solid lines and shaded bands indicate the means and SDs from the data from five samples. (E) The functional block diagram of the magnetic robot with magnetic actuation systems (the electromagnet array or mobile coil system). (F) Real-time digital twin visualization demonstrating precise posture replication and monitoring.

Fig. 4.. Digital twin–enabled embodied intelligence for autonomous environmental exploration and perturbation correction.

(A) Mobile coil system implementation for automated obstacle avoidance on water surface through real-time sensing (B) Digital twin visualization of autonomous narrow passage navigation, showing real-time posture adjustment capability. (C) Autonomous decision-making demonstration: digital twin recording of U-turn execution when encountering impassable obstacles. (D) Electromagnet array configuration for dynamic perturbation correction. (E) Position and orientation restoration following external disturbances using the perturbation correction program. (F) Comparative yaw angle analysis with and without perturbation correction program activation. (G) Pitch and roll angle stability maintenance during perturbation correction.