NEWS AND VIEWS

Robotic collectives inspired by biological cells

A robotic system has been demonstrated in which the random motion of individual components leads to deterministic behaviour, much as occurs in living systems. Environmental and medical applications could follow.
Metin Sitti is in the Physical Intelligence Department, Max Planck Institute for Intelligent Systems, 70569 Stuttgart, Germany, and at the School of Medicine and the School of Engineering, Koç University, Istanbul, Turkey.
Contact

Search for this author in:

In biological systems, large-scale behaviour can be achieved by the collective coupling and coordination of stochastically (randomly) moving small-scale components. For example, living cells aggregate and migrate collectively during the healing of wounds and when cancer spreads. Inspired by these biological mechanisms, in a paper in Nature, Li et al.1 report a collective robotic system in which deterministic locomotion is a result of the stochastic movement of many loosely coupled, disc-shaped components. The results show that stochasticity offers a promising approach to developing large-scale, collective robotic systems that exhibit robust deterministic behaviour.

In Li and colleagues’ system, the disc-shaped components cannot move independently of one another and cannot be manipulated individually. Moreover, each component can move only by oscillating along its radius, by expanding and contracting. The authors refer to this minimalistic approach as “particle robotics”. In the absence of an external stimulus, the system can move only randomly. However, when the components are programmed to adjust their diameter in response to a varying environmental signal, there is collective locomotion towards the source of the signal.

Li et al. carry out experiments in which the particle-robotics system contains up to two dozen components, and simulations in which the system has as many as 100,000 components. The diameter of each component varies from 15.5 to 23.5 centimetres during the oscillations. The authors show that the system can achieve robust locomotion and object transport, as well as light-directed movement and obstacle avoidance (Fig. 1). Remarkably, they find that locomotion can be maintained even when 20% of the components malfunction, which highlights the robustness of the particle-robotics approach to individual component failures.

Figure 1 | An innovative collective robotic system. a–c, Li et al.1 report a robotic system that is composed of many loosely coupled, randomly moving, centimetre-scale components. Each component can move only by oscillating along its radius, by expanding and contracting. The colour of the components represents their diameter, from minimum (green) to maximum (blue), during such oscillations. Malfunctioning components, which are used to test the robustness of the system, are depicted in maroon. The authors show that their system can exhibit deterministic locomotion towards an environmental signal — such as a light source — over time, while avoiding obstacles. (Adapted from Fig. 1f of ref. 1.)

Previous studies have mainly considered components that can move independently of one another, can be manipulated individually, and that are based on relatively complex, deterministic designs25. Most of the previously reported collective robotic systems have limited flexibility in terms of allowable configurations, whereas those that are amorphous typically contain components with limited scalability. Moreover, many of these systems require some level of centralized control, which further restricts their capabilities and scalability.

In this regard, Li and colleagues’ particle-robotics approach provides a promising alternative to other methods. In addition to being inspired by biological systems, the technique is motivated by statistical-physics phenomena, in which the global statistical behaviour of a large number of stochastic components can be modelled and controlled without the need to track each component. As a result, the approach has substantial advantages over other methods, especially when scaling up the number of components and scaling down the size of each one. Such scaling will be required for many future potential applications of collective robotic systems in exploration, construction and medicine.

Nevertheless, the authors’ system has some drawbacks. First, if there is no environmental-signal gradient at the location of the aggregate of components, the system cannot move towards the source of the signal. Second, the components need to start from manually configured positions, because they cannot move independently to engage each other. Third, the experimentally demonstrated components are limited in number and are relatively slow and large; in the near future, the system should be extended to a larger number of components that are faster and much smaller (perhaps even down to the micrometre scale). And fourth, this technique is not suitable for tasks such as directed self-assembly and self-organization into complex, prespecified geometries, because of the stochastic nature of the aggregate and the random placement and coupling of the components.

Because of advances in small-scale robotics, it is possible to design and fabricate large numbers of stochastic or deterministic components that can exhibit collective and swarming behaviour6 similar to that of the particle-robotics system. In the past few years, mobile microscale robotic swarms that have well-defined collective behaviour have been produced by engineering magnetic interactions between individual units. In general, the main strategy for controlling such swarms relies on the response of the units to remotely controlled global fields, such as magnetic fields7,8. Although it is then difficult to deal with each unit individually and locally, collective coupled interactions between the units can be globally controlled, resulting in programmable local interactions, self-assembly and collective behaviour. This method has been used to attain collective two-dimensional assembly, disassembly and manipulation of synthetic microrobotic swarms at the interface between air and water9.

Li and colleagues’ particle-robotics system, and most other collective robotic systems, work mainly in two dimensions. Extending such systems to three dimensions, with more-complex locomotive behaviour of components and their aggregates on surfaces or inside fluids, would increase their possible future applications. However, going to three dimensions would bring many hardware-design challenges for robust locomotion, aggregate stability, reversible and programmable component-attachment methods, miniaturization and control.

In the near future, it will be crucial to demonstrate potential high-impact engineering and medical applications of such collective robotic systems that would be impossible using other techniques. For example, swarms of stochastic bacterium-driven microrobotic swimmers could use the particle-robotics approach to deliver drugs to targeted, hard-to-reach regions inside the human body. Such swarms might, for example, be directed by the chemical gradients, oxygen gradients or changes in pH of cancerous-tissue environments10. Indeed, many studies11,12 have already shown that collective bacterium-driven microrobotic swarms have potential applications in targeted drug delivery, medical diagnostics and environmental sensing.

Nature 567, 314-315 (2019)

doi: 10.1038/d41586-019-00839-x

References

  1. 1.

    Li, S. et al. Nature 567, 361–365 (2019).

  2. 2.

    Chen, J., Gauci, M., Li, W., Kolling, A. & Gross, R. IEEE Trans. Robot. 31, 307–321 (2015).

  3. 3.

    Butler, Z. & Rus, D. Int. J. Robot. Res. 22, 699–715 (2003).

  4. 4.

    Gross, R., Bonani, M., Mondada, F. & Dorigo, M. IEEE Trans. Robot. 22, 1115–1130 (2006).

  5. 5.

    Yim, M. et al. IEEE Robot. Autom. Mag. 14, 43–52 (2007).

  6. 6.

    Sitti, M. Mobile Microrobotics (MIT Press, 2017).

  7. 7.

    Yigit, B., Alapan, Y. & Sitti, M. Adv. Sci. 2019, 1801837 (2019).

  8. 8.

    Yu, J., Wang, B., Du, X., Wang, Q. & Zhang, L. Nature Commun. 9, 3260 (2018).

  9. 9.

    Wang, W., Giltinan, J., Zakharchenko, S. & Sitti, M. Sci. Adv. 3, e1602522 (2017).

  10. 10.

    Park, B.-W., Zhuang, J., Yasa, O. & Sitti, M. ACS Nano 11, 8910–8923 (2017).

  11. 11.

    Alapan, Y. et al. Sci. Robot. 3, eaar4423 (2018).

  12. 12.

    Felfoul, O. et al. Nature Nanotechnol. 11, 941–947 (2016).

Download references

Nature Briefing

Sign up for the daily Nature Briefing email newsletter

Stay up to date with what matters in science and why, handpicked from Nature and other publications worldwide.

Sign Up