It is extremely rare that a robotics advance, in one fell swoop, shrinks the size of a platform by a factor of 10,000 times, without any compromise to its fundamental capabilities. And yet, that is exactly what has been achieved by this joint University of Pennsylvania and University of Michigan team: fully programmable, autonomous robots at the scale of bacteria, each with the ability to move, sense, compute, and act untethered for months in liquid environments.
1. Breaking the Sub-Millimeter Barrier
For a long time, miniaturization in robotics appeared to plateau above one millimeter in size-the physics at the microscale simply get in the way. Drag and viscosity start to dominate other forces at these scales, rendering conventional limbs and propellers useless. “If you’re small enough, pushing on water is like pushing through tar,” said Marc Miskin, assistant professor at Penn Engineering. Added to that, fragility and fabrication limits made self-propelled, autonomous motion near impossible without being tethered to external control systems.
2. Electrokinetic Propulsion Without Moving Parts
The solution took the form of an electrokinetic propulsion system that avoids mechanical parts altogether. The robots use oppositely biased electrodes, through which they pass a current, to create an electric field. This electric field moves the ions in the surrounding fluid. The moving ions take water molecules with them, creating a flow that pushes the robot. “It’s like the robot is in a river and the river is moving, but the robot is also moving the river,” Miskin explained. This technique promises to be durable for several months and can be mass-produced using lithographic techniques, therefore capable of large-scale fabrication at extremely low cost.
3. Ultra-Low-Power Computing at Microscopic Scale
Autonomy necessitates on-board computation, which must be performed within a stringent power budget: approximately 75 nanowatts, more than 100,000 times smaller than for a smartwatch. In this regard, the team from David Blaauw at University of Michigan custom-designed circuits with subthreshold digital logic to meet these constraints. Processor and memory, photovoltaic cells, and sensors fit within a footprint of roughly 200 × 300 × 50 micrometers.
4. Specialized Instruction Set Architecture
Given that there is only a few hundred bits of memory, the team developed a complex instruction set adapted to microrobotic tasks. Compressing commands like “sense temperature” or “move for N cycles” into single operations represents what normally requires dozens of instructions. This architecture provides digitally defined algorithms, closed-loop control, and reprogrammability after fabrication within the constrained resources of the robot.
5. Optical Power and Programming
The surface of each robot is mostly covered with microscopic solar cells that harvest light for energy and data transfer. Pulses of light program the robots; each has a unique address enabling directed instructions. This allows many robots to share the same environment and carry out different roles in coordinated tasks-a key factor in scalable swarm operations.
6. Integrated Sensing and Adaptive Behaviour
For this demonstration capability, temperature sensing accurate to within 0.3°C was chosen. In experiments, robots measured local temperatures, recorded the information, and transmitted it by modulating their motion in a “dance” pattern-a range of physical encoding reminiscent of honeybee communication. When exposed to a temperature gradient, the robots altered propulsion action to climb toward warmer regions via autonomous navigation without external control.
7. Micro-Scale Physics Meets System Integration
Integration on this scale required both fabrication processes compatible with CMOS foundry and sensitive post-processing steps to release the robots into solution. Platinum patterning of electrokinetic actuators, circuit encapsulation in protective oxide, and thinning the entire assembly to 50 micrometers were done before release. Typical yields were in excess of 50%, with the resulting devices showing repeatable performance across a diversity of fluid environments.
8. Possible Applications and Future Directions
Operating at the scale of microorganisms opens avenues in medicine, biology, and manufacturing. The robots could monitor individual cell health, perform targeted sensing in microfluidic chambers, or assemble microscale structures. Future iterations may expand memory by 100-fold, increase propulsion speed tenfold, or integrate additional sensors for biochemical detection. The combination of onboard computation, reprogrammability, and low-cost fabrication positions these robots as a general-purpose platform for autonomous operation in environments too small or complex for conventional machines.
“This is really just the first chapter,” Miskin said. “We’ve shown that you can put a brain, a sensor, and a motor into something almost too small to see, and have it survive and work for months.”