In microrobot design, physical, biological, computational, and human intelligence each provide distinct capabilities. Their integration forms the foundation of cross-domain intelligence.
| Domain | Role | Key Mechanisms | Key Challenges | Representative Tasks |
|---|---|---|---|---|
| Physical Intelligence (PI) | Material and structure-driven autonomy | 1. Stimuli-responsive materials (e.g., hydrogels, liquid crystals), magnetoelastic composites; 2. Bioinspired geometries (e.g., helical structures) | Limited selectivity, difficulty in multifunctional integration | pH/targeted drug delivery, intravascular navigation |
| Biological Intelligence (BI) | Integration of living cells/organisms for actuation, sensing, and adaptability | Biohybrid robots using iPSC, neutrophils, E. coli, or platelets | Biological variability, short lifespan, limited controllability in vivo | Tumor-homing targeted therapy, chemotaxis-based cargo transport |
| Computational Intelligence (CI) | Algorithm-driven perception, learning, and control | CNN for perception, RL/IL for navigation, digital twins for simulation and optimization | Limited onboard computing, data scarcity, high safety requirements | Image-based depth/pose estimation, closed-loop navigation |
| Human Intelligence (HI) | High-level cognitive guidance and expert supervision | Teleoperation, haptic feedback, AR/VR interfaces, shared control | Latency and interface limitations, limited situational awareness in 3D | VR-guided vascular navigation, remote micromanipulation |
Note: Table content is based on the summary of intelligence domains in the paper's Table 1.
Definition and Significance: Physical intelligence directly encodes sensing, actuation, control, memory, logic, and adaptation capabilities into the physical entity (materials and structures) of microrobots without relying on a central digital brain or external controllers.
Mechanism Examples: Hydrogel microrobots release drugs in acidic tumor microenvironments; heat-triggered untethered microgrippers close at body temperature to complete tissue biopsy; magnetoelastic soft composites achieve deformation and motion under magnetic field driving; helical structures improve propulsion efficiency in viscous media.
Challenges and Prospects: Currently, PI has limited functional selectivity and difficulty integrating multiple functions within constrained volumes. Future needs include developing multi-responsive degradable materials, reconfigurable structures, and computational optimization design.
Definition and Significance: Biological intelligence endows robots with inherent biological actuation, sensing, and adaptation capabilities by introducing living cells or microorganisms.
Mechanism Examples: iPSC-driven microrobots can utilize tumor-homing capabilities to deliver nanomedicines; neutrophil carriers use chemotaxis for targeted transport; E. coli-driven robots can navigate in hypoxic regions; platelet-based systems achieve targeted localization through receptor-mediated adhesion.
Challenges and Prospects: BI systems are greatly affected by the biological internal environment and individual differences, have short lifespans, and are difficult to operate stably for extended periods. Future approaches will programmatically control cell behavior through genetic engineering and synthetic biology, combined with closed-loop sensing to improve reliability.
Definition and Significance: Computational intelligence provides algorithm-driven perception, learning, and control, enabling microrobots to sense, plan, and make decisions in dynamic environments.
Mechanism Examples: Convolutional neural networks for image perception, reinforcement learning and imitation learning for navigation and strategy learning, digital twins for virtual-physical integration simulation and control.
Challenges and Prospects: Constrained by size limitations, onboard computing resources are scarce; real data is limited, and medical safety requirements are high. Future needs include developing efficient models, methods that integrate data-driven approaches with physical constraints, and verification mechanisms to ensure clinical safety.
Definition and Significance: Human intelligence supplements knowledge gaps in the system through expert guidance and cognitive intervention, ensuring safety in uncertain environments.
Mechanism Examples: Remote manipulation and haptic feedback enable doctors to finely manipulate microrobots; AR/VR interfaces provide intuitive immersive experiences for operations; shared control frameworks combine machine autonomous decision-making with human commands.
Challenges and Prospects: Interface latency and insufficient 3D perception limit operational precision. Future development needs include real-time imaging, ergonomic interfaces, and intelligent assistant algorithms to improve human-robot collaboration efficiency.