Insect-scale robots get a brain upgrade through embodied intelligence

Researchers at the University of Macau say tiny robots need a new approach to autonomy: let their bodies, materials and environment do more of the work. Their new perspective in eScience outlines how embodied intelligence could help millimeter- and centimeter-scale robots operate in search and rescue, agriculture and other hard-to-reach settings.

Why it matters: - Insect-scale robots could improve environmental monitoring, disaster response, precision agriculture, infrastructure inspection and even some medical tasks. - Their tiny size makes autonomy hard because payload, energy storage and computing power are all extremely limited. - The paper argues that better performance will come from co-designing hardware and behavior, not just shrinking conventional processors.

What happened: - Researchers from the University of Macau’s Department of Electromechanical Engineering and Centre for Artificial Intelligence and Robotics published a perspective paper in eScience. - The paper appears in Volume 6, Issue 4, July 2026. - The article is titled around embodied intelligence for insect-scale robots and is indexed with DOI 10.1016/j.esci.2026.100549. - The original source is available through the published paper.

The details: - The authors frame embodied intelligence as a way for smart, adaptive behavior to emerge from a robot’s physical form, materials and interaction with the environment. - The paper points to bio-inspired crawlers, swimmers and fliers as evidence that locomotion at small scale has advanced quickly. - Soft materials such as shape-memory alloys, dielectric elastomers and liquid crystal elastomers can support large deformation and stimulus-responsive behavior. - Chemical fuels and biofuel cells are among the power approaches proposed for tiny robots. - The authors say these advances have not yet solved autonomous operation in unstructured, unpredictable environments. - At the physical level, the paper promotes “physical intelligence,” where smart materials and adaptive structures reduce the need for computation. - Origami and kirigami structures can turn flat sheets into complex 3D forms. - Multi-stable mechanisms can store elastic energy and support self-adjusting behavior without continuous external input. - Bionic bodies with passive, compliant joints can move over rough terrain without real-time feedback control. - On the computing side, the paper describes a scalability paradox: conventional AI needs hardware and power beyond insect-scale limits, while cloud processing adds too much latency. - TinyML is presented as one option, with machine-learning models compressed onto microcontrollers using less than 100 KB of memory and sub-milliwatt power budgets. - Bio-inspired algorithms are another option. - Central pattern generators and Braitenberg-style sensor-motor couplings can produce behaviors such as phototaxis and obstacle avoidance with little or no training data. - At the swarm level, the paper says collective intelligence can add robustness and scale, but communication remains a bottleneck. - Conventional wireless systems do not fit large-scale swarms, and even Kilobot can take hours to form simple shapes.

Between the lines: - The paper is arguing for a shift in what “intelligence” means at very small scales. - That view challenges a common robotics instinct: add more compute to get more autonomy. - The authors are effectively saying that tiny robots may need morphology, mechanics and control strategies that encode behavior directly into the machine. - The broader implication is that progress in this field will depend on combining materials science, mechanics, computer science and biology rather than optimizing any one discipline alone. - The paper also flags ethics and public policy as part of the transition from lab prototypes to real-world use. - Privacy, responsible deployment and lifecycle management will matter as these robots move into practical settings.

What’s next: - The authors call for interdisciplinary collaboration to turn laboratory prototypes into practical intelligent systems. - Future work will likely focus on matching embodied design with low-power computation and swarm communication methods. - The paper suggests that the next major milestone is autonomous operation in messy, real-world environments rather than controlled demonstrations. - Funding for the work came from the National Natural Science Foundation of China and the Science and Technology Development Fund, Macau SAR.