Noncontact manipulation in liquid environments holds significant applications in micro/nanofluidics, microassembly, micromanufacturing, and microrobotics. Achieving compatibility in manipulating both sedimented and floating objects, as well as independently and synergistically manipulating multiple targets, remains a significant challenge. Here, a noncontact manipulator is developed for both sedimented and floating objects using laser-induced thermocapillary convection. Various strategies are proposed based on the distinct responses of sedimented and floating objects. Predefined scanning and “checkpoint” methods facilitate accurate movements of individual and multiple particles, respectively. Ultrafast programmed scanning and laser multiplexing enable independent manipulation and high-throughput ordered distribution of multiple particles. At the air–liquid interface, “laser cage” and “laser wall” are proposed to serve as effective tools for manipulating floating objects, especially with vision-based closed-loop control. Methods and strategies here do not rely on specific features of targets, solvents, and substrates. Multiple examples, including complex path replication, maze traversal, and precise assembly and disassembly, are demonstrated to validate the feasibility of this manipulator. This work provides a versatile platform and a novel methodology for noncontact manipulation in liquid.

As a classical type of smart materials, shape memory alloys (SMAs) are of high energy density, light weight, and low actuating voltage, and therefore are of great potential to be used as actuators for robots. Major challenges in controlling an SMA-driven soft robot are the limited bandwidth and in cases with external loads. Active cooling has been demonstrated to dramatically increase its bandwidth, but external load may cause severe inaccuracy in the system modeling. Controllers that do not rely on accurate modeling of the SMA-driven soft robot is essential. In this article, we designed an elastomeric soft robotic tentacle actuated by three pieces of SMA springs with both active heating (Joule heating) and active cooling (compressed air). We proposed a multi-input-multi-output controller that directly uses the heating and cooling states of the three SMAs to control the tentacle's bending posture in three-dimensional (3-D) space. The successful implementation of the controller is attributed to a novel dual-channel control algorithm that integrates the bending motion control and swing motion control, and a state-machine controller for coordinating the three SMAs' actuations to achieve robust swing motion control. The system with such hardware and control algorithm was capable of performing bending motions with maximum actuating speed$ > $90$^\circ /$s, deactuating speed$ > $25$^\circ /$s, closed-loop motions with rapidity (6–41$^\circ /$s for heating, 4–19$^\circ /$s for cooling), accuracy (steady-state error$ < $0.1$^\circ$for no load, 0.13$\%$of the full range; steady-state error$ < $1.2$^\circ$with load of 1 bodyweight, 1.6$\%$of the full range), and load-bearing capability (dynamic load: 1 bodyweight, static load: 8.8 bodyweights). Besides, the tentacle achieved efficient motion tracking by coordinating the bending and swing motion through the controller, for both predefined trajectories and random trajectories. We demonstrated a remotely controlled 360$^\circ$image scanning of a room using our proposed robotic tentacle equipped with a camera at its top end to intuitively show its performances. We believe this work will advance the design and control of SMA-driven soft continuum robots for potential uses in surgery and explorations of unknown areas.