MICoBot: Multiscale Robotic Innovations

• MICoBot is a multifaceted term describing diverse microrobotic systems that integrate sensing, actuation, and onboard computing across microfluidic, modular, and sub-millimeter scales.
• One implementation involves a magnetically levitated microrobotic mixer in a closed microfluidic chip that achieved up to 80.37% mixing efficiency under optimized rpm and flow conditions.
• Other variants include self-folding smartlet cubes for energy harvesting and docking, as well as lithographically fabricated robots with sub-threshold digital logic for ultra-low power autonomous computing.

Searching arXiv for papers using the term “MICoBot” and closely related titles to ground the article in published work. MICoBot is a reused research name rather than a single standardized robot. In recent arXiv literature, it denotes several distinct systems: a magnetically levitated microrobotic mixer for closed microfluidic chips, modular electronic microrobots built as self-folding micro-origami cubes or “smartlets,” a lithographically fabricated sub-millimeter robot that can “sense, think, act, and compute” on-board, and, in a different subfield, a Mixed-Initiative Collaborative roBot for long-horizon human-robot manipulation (Yılmaz et al., 2023, Lee et al., 2024, Bandari et al., 24 Aug 2025, Lassiter et al., 29 Mar 2025, Yu et al., 7 Aug 2025). The term therefore functions less as a single platform name than as a label attached to multiple efforts that emphasize integrated sensing, computation, actuation, and task-level autonomy under strong physical or interaction constraints.

1. Scope of the term and antecedents

Within the cited record, MICoBot spans at least three scales: microfluidic microrobotics, sub-millimeter electronic microrobots, and macroscale mobile or collaborative robots. A concise disambiguation is useful.

arXiv id	MICoBot usage	Scale/domain
(Yılmaz et al., 2023)	Magnetically levitated microrobotic mixer	Closed microfluidics
(Lee et al., 2024)	3D modular microrobots / smartlets	$< 1\ \text{mm}^3$ modular microrobotics
(Bandari et al., 24 Aug 2025)	Modular electronic microrobot smartlet with 2D locomotion	$\le 1$ mm cube platform
(Lassiter et al., 29 Mar 2025)	“Microscopic Robots That Sense, Think, Act, and Compute”	Sub-millimeter CMOS microrobot
(Yu et al., 7 Aug 2025)	Mixed-Initiative Collaborative roBot	Human-robot collaborative manipulation
(Iqbal et al., 4 Aug 2025)	Autonomous microplastics surveying robot	Beach mobile manipulator

A useful conceptual precursor is the earlier review of microscopic biomedical robots that emphasized operation in fluids at low Reynolds number, diffusion-limited sensing, Brownian motion, and millisecond-scale distributed control decisions [0611111]. That work did not use the MICoBot name, but it framed several constraints that later MICoBot-labeled microrobotic systems still confront: limited power, restricted sensing bandwidth, fluid-dynamic nonidealities, and the trade-off among speed, accuracy, and resource use.

A common misconception is that MICoBot always refers to a micrometer-scale swimmer or manipulator. The literature represented here does not support that reading. Instead, the label has been attached to systems ranging from sealed-chip micromixers to household collaborative robots.

2. MICoBot as a magnetically levitated microrobotic mixer

In one usage, MICoBot denotes a magnetically levitated microrobotic mixer integrated into a PMMA microfluidic chip for mixing NaOH solution and thymolphthalein indicator under laminar flow conditions (Yılmaz et al., 2023). The chip dimensions are $$24\ \text{mm} \times 40\ \text{mm} \times 2\ \text{mm}$$, with two inlet channels, one mixing chamber, and one outlet channel in a Y-type arrangement. The actuation concept is non-contact: one microrobot is attached to a stepper motor rotor as the external actuator, and a second microrobot with embedded NdFeB permanent magnets (N52) is placed inside the mixing chamber as the mixing robot. Magnetic coupling drives the internal robot to rotate or levitate, producing convective disturbance that supplements diffusion-limited microfluidic transport.

The flow model is explicitly laminar, incompressible, and Newtonian. The paper states the Reynolds number as

$Re = \frac{\rho U D_h}{\mu},$

and evaluates mixing through grayscale image analysis. After image capture, grayscale conversion, and normalization, the mixed state is estimated from the mean intensity along a straight line through the channel exit, using the reported mixing index

$MI = \frac{\sum_{i=1}^{n} C_i}{n}.$

Here, concentration is represented through grayscale intensity, so greater spatial uniformity corresponds to better mixing.

The tested rotational speeds were $100$, $250$, $500$, and $800\ \text{rpm}$, and the five flow rates were $0.5$, $\le 1$0, $\le 1$1, $\le 1$2, and $\le 1$3, producing 25 tests. The best reported performance was a mixing efficiency of $\le 1$4 at $\le 1$5 and $\le 1$6. Efficiency generally increased with flow rate, but the speed dependence was non-monotonic: $\le 1$7 often produced insufficient torque for stable levitation or rotation, $\le 1$8 gave the best overall performance, and $\le 1$9 reduced stability and lowered efficiency across all flow rates. The paper therefore presents MICoBot not as a generic “faster is better” mixer, but as a system with a clear actuation optimum bounded by magnetic alignment and dynamic stability.

The significance of this MICoBot usage lies in closed-chip, non-contact mixing. The reported implications include microfluidic chemical synthesis, nanoparticle synthesis, encapsulation, lower reagent consumption, and safer handling of hazardous chemicals. Its principal limitations are equally explicit: image-based mixing is an optical proxy rather than a direct concentration field measurement, and the reported conclusions assume laminar, incompressible, Newtonian flow.

3. MICoBot as a modular electronic smartlet

A second MICoBot lineage consists of modular electronic microrobots built as self-folding micro-origami cubes, called smartlets, that integrate energy harvesting, sensing, computation, communication, actuation, and docking in a three-dimensional architecture (Lee et al., 2024). In this platform, a planar polymer stack comprising a sacrificial layer, a photo-patternable hydrogel hinge or rolling layer, a polyimide rigid layer, and an SU-8 support layer self-folds and self-rolls into a cube of roughly 1 mm scale. The resulting smartlet carries edge-mounted rolled micro-organic solar cells, interior chambers for buoyancy control, custom CMOS chiplets, and micro-LEDs for optical communication. The 2024 paper reports eight rolled micro-organic solar cells on the eight cube edges, with eight devices in series giving $$24\ \text{mm} \times 40\ \text{mm} \times 2\ \text{mm}$$0 up to $$24\ \text{mm} \times 40\ \text{mm} \times 2\ \text{mm}$$1, $$24\ \text{mm} \times 40\ \text{mm} \times 2\ \text{mm}$$2, $$24\ \text{mm} \times 40\ \text{mm} \times 2\ \text{mm}$$3, and output power about $$24\ \text{mm} \times 40\ \text{mm} \times 2\ \text{mm}$$4. Communication is optical and local: the green LED can send data from $$24\ \text{mm} \times 40\ \text{mm} \times 2\ \text{mm}$$5 to $$24\ \text{mm} \times 40\ \text{mm} \times 2\ \text{mm}$$6 over water at separations under $$24\ \text{mm} \times 40\ \text{mm} \times 2\ \text{mm}$$7, enabling start and stop commands and program-triggered actuation.

In that earlier smartlet formulation, locomotion is based on electrochemical bubble generation inside the cube, producing programmed buoyancy switching: bubble generation begins, gas accumulates, the smartlet rises, actuation stops, the bubbles dissolve, and the module sinks. The same paper also reports face-selective self-assembly through hydrophobic and hydrophilic barcode-like surface regions, allowing docking into chains, square arrays, and partially guided assemblies spelling letters such as T, U, and C (Lee et al., 2024).

A later extension uses the MICoBot name for a modular electronic microrobot smartlet that adds 2D locomotion on wet surfaces with sensor-program steered navigation and selective docking (Bandari et al., 24 Aug 2025). Here the main body is a hollow cube of about $$24\ \text{mm} \times 40\ \text{mm} \times 2\ \text{mm}$$8, open on one face, with a smaller version of $$24\ \text{mm} \times 40\ \text{mm} \times 2\ \text{mm}$$9 edge length also noted and a claimed path toward $Re = \frac{\rho U D_h}{\mu},$0. The on-board controller is a custom CMOS microchiplet, or “lablet,” of dimensions $Re = \frac{\rho U D_h}{\mu},$1, carrying a 58-bit program and using two contact pads as a 1-bit digital input interface with sensors. Power comes from ambient light harvested by custom submersible organic photovoltaic “Swiss-rolls,” with a $Re = \frac{\rho U D_h}{\mu},$2 capacitor used for storage and conditioning. Integrated photodetectors provide thresholded digital sensing, with rise and fall times of about $Re = \frac{\rho U D_h}{\mu},$3 and $Re = \frac{\rho U D_h}{\mu},$4, respectively.

Locomotion is driven by three independently addressable bubble-generating electrodes, BGE1, BGE2, and BGE3. The mechanism is ratcheting motion on a thin water film about $Re = \frac{\rho U D_h}{\mu},$5 deep: electrolysis generates bubbles, bubbles accumulate asymmetrically on one face, pressure lifts one side of the cube, and bubble escape drives forward displacement. The paper estimates bubble diameters of about $Re = \frac{\rho U D_h}{\mu},$6–$Re = \frac{\rho U D_h}{\mu},$7, uses the Laplace-pressure relation

$Re = \frac{\rho U D_h}{\mu},$8

and reports approximately $Re = \frac{\rho U D_h}{\mu},$9 displacement per tilting cycle at about $MI = \frac{\sum_{i=1}^{n} C_i}{n}.$0 cycles/s, corresponding to speed around $MI = \frac{\sum_{i=1}^{n} C_i}{n}.$1, measured at $MI = \frac{\sum_{i=1}^{n} C_i}{n}.$2. A bright-region sensing experiment shows that threshold crossing can switch the active face and produce a $MI = \frac{\sum_{i=1}^{n} C_i}{n}.$3 turn. Selective docking is achieved by hydrophilic and hydrophobic surface patterns; hydrophilic-hydrophilic faces dock, whereas hydrophobic-hydrophilic pairs repel or fail to dock (Bandari et al., 24 Aug 2025).

Taken together, these smartlet papers define MICoBot as a route toward autonomous modular microrobotics. The common architectural move is three-dimensional surface enrichment: edges for light harvesting, faces for sensors and docking, interior volume for actuation and control. A plausible implication is that the name MICoBot, in this lineage, is attached less to a single actuation mode than to a strategy of heterogeneous integration at sub-millimeter scale.

4. MICoBot as a lithographically fabricated “sense-think-act” microscopic robot

A third, more computation-centric usage presents MICoBot as a sub-millimeter electronic robot fabricated in a 55 nm CMOS process and explicitly designed to “sense, think, act, and compute” on-board (Lassiter et al., 29 Mar 2025). One reported form factor occupies about $MI = \frac{\sum_{i=1}^{n} C_i}{n}.$4, with another version at $MI = \frac{\sum_{i=1}^{n} C_i}{n}.$5 body width. The paper places strong emphasis on the power budget, approximately $MI = \frac{\sum_{i=1}^{n} C_i}{n}.$6, and uses sub-threshold digital logic, low-leakage semiconductor technology, and compact instruction encoding to fit useful autonomy into that envelope.

The computational substrate is a custom onboard processor with an 11-bit CISC-ISA, a 32-entry 11-bit instruction memory, a 4-register file with four 8-bit registers, and a 16-entry 8-bit data memory. Standard instructions include arithmetic, logic, shifts, branches, and load-store operations, but the architecture also introduces robot-specific instructions: mot repeats a programmed actuator sequence for $MI = \frac{\sum_{i=1}^{n} C_i}{n}.$7 cycles, ts senses temperature and stores the result, and wav Manchester-encodes a value and transmits it through actuator motion. The system’s memory remains highly constrained—roughly hundreds of bits overall—so the paper uses progressive, multi-step programming in which an initialization program is loaded first and a task program is then sent optically.

Sensing is demonstrated through a temperature sensor with $MI = \frac{\sum_{i=1}^{n} C_i}{n}.$8 resolution in a volume below $MI = \frac{\sum_{i=1}^{n} C_i}{n}.$9. Communication is optical and passcode-gated: a power LED provides harvested energy, a communication LED transmits a passcode sequence, and only after passcode recognition does the robot accept incoming bits into instruction memory. The paper reports both a global passcode and a type-specific passcode, enabling selective programming of subsets of robots. Locomotion uses electrokinetic propulsion through patterned platinum electrodes, with low current of less than $100$0, voltage about