Safety of human-robot interaction through tactile sensors and peripersonal space representations

05/18/2018 ∙ by Petr Švarný, et al. ∙ Czech Technical University in Prague 0

Human-robot collaboration including close physical human-robot interaction (pHRI) is a current trend in industry and also science. The safety guidelines prescribe two modes of safety: (i) power and force limitation and (ii) speed and separation monitoring. We examine the potential of robots equipped with artificial sensitive skin and a protective safety zone around it (peripersonal space) to safe pHRI.



There are no comments yet.


page 2

This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.

1 Introduction

The combination of words safety and robotics often incites images of a machine uprising in the minds of laymen. Contemporary robotics, however, faces a great deal of challenges connected to even mundane interaction scenarios between robots and humans.

2 Standardization

The overall safety in physical human-robot interaction (pHRI) is subject to many standards. These start with the general machinery standards like ISO 12100, ISO 13849, followed by specific robot standards as ISO 10218. The speed of robotics evolution makes standardization very difficult. The newest standard in preparation ISO/TS 15066 mirrors the trend of collaborative robotics, but still has a lot of discussion ahead before it can become an accepted standard Haddadin (2015)Jacobs et al. (2018).

3 Collaborative operation

Robots have to adhere to all the mechanical safety standards as any other industrial machinery. However, as opposed to classical machines, robots can have complex behavior while interacting with people.

The ISO 10218 and ISO/TS 15066 specify four types of safe pHRI:

  1. Safety-rated monitored stop

  2. Hand guiding

  3. Power and force limiting by inherent design or control

  4. Speed and separation monitoring

In the first two regimes, there is no simultaneous autonomous movement of the robot and the human collaborator allowed: in 1), the robot will stop whenever the human enters the workspace; in 2), the robot operates in a specific hand-guiding (kinesthetic teaching) mode and does not execute any independent movements. The other two regimes, on the other hand, constitute the real challenge.

Figure 1: Robot in a monitored space.

3.1 Power and force limiting

Power and force limiting allows physical contacts with a moving robot but they need to be within human body part specific limits on force, pressure, and energy. Example of the safety foundation on the robot side is a lightweight structure. The perception of collisions leads to appropriate reactions (e.g., (Magrini et al., 2015)). A recent survey on this post-impact interaction control is (Haddadin et al., 2017).

3.2 Speed and separation monitoring

Speed and separation monitoring deals with pre-impact

interaction. It demands reliable estimation of distances between robots and humans. Proper estimation allows the alteration of the robots behavior in order to maintain the minimal separation distance between the operator and the robot that cannot be crossed. However, light curtains or safety-rated scanners (e.g., SafetyEYE

111 are very coarse (monitor 2D or 3D zones) whereas sensor with higher resolution (e.g., cameras or RGB-D sensors) from which also human skeleton can be extracted are currently not safety-rated Flacco et al. (2015); Nguyen et al. (2018).

Figure 2: Schematics of a receptive field that is part of the peripersonal space of the iCub robot (Roncone et al., 2016).

4 Artificial skins and peripersonal space

Our own research uses robots with pressure-sensitive skins, like the iCub. These can be exploited for contact detection and response but also for calibration of the safety margin through visuo-tactile associations (see Fig. 2) and (Roncone et al., 2016)

). Alternatively, the safety margin can rely on distal sensing using cameras or RGB-D sensors and human skeleton extraction by convolutional neural networks. The availability of safety-rated human keypoint extraction or at least point cloud detection would dramatically expand the possibilities of human-robot collaboration in the speed and separation monitoring regime.

5 Conclusion

Safe pHRI is a dynamically evolving field with some borders set by industry standards but with a vivid discussion about best practices.

Petr Švarný was supported by the Grant Agency of the Czech Technical University in Prague, grant No. SGS18/138/OHK3/2T/13. Matěj Hoffmann was supported by the Czech Science Foundation under Project GA17-15697Y.


  • ISO (2010) (2010). ISO 12100 Safety of machinery – General principles for design – Risk assessment and risk reduction. Standard, International Organization for Standardization, Geneva, CH.
  • ISO (2011) (2011). ISO 10218 Robots and robotic devices – Safety requirements for industrial robots. Standard, International Organization for Standardization, Geneva, CH.
  • ISO (2015) (2015). ISO 13849 Safety of machinery – Safety-related parts of control systems. Standard, International Organization for Standardization, Geneva, CH.
  • ISO (2016) (2016). ISO/TS 15066 Robots and robotic devices – Collaborative robots. Standard, International Organization for Standardization, Geneva, CH.
  • Flacco et al. (2015) Flacco, F., Kroeger, T., De Luca, A. and Khatib, O. (2015). A depth space approach for evaluating distance to objects. Journal of Intelligent & Robotic Systems, 80(1):7–22.
  • Haddadin (2015) Haddadin, S. (2015). Physical safety in robotics. V Formal modeling and verification of cyber-physical systems, str. 249–271. Springer.
  • Haddadin et al. (2017) Haddadin, S., De Luca, A. and Albu-Schäffer, A. (2017). Robot collisions: A survey on detection, isolation, and identification. IEEE Transactions on Robotics, 33(6):1292–1312.
  • Jacobs et al. (2018) Jacobs, T., Veneman, J., Virk, G. S. and Haidegger, T. (2018). The flourishing landscape of robot standardization [industrial activities]. IEEE Robotics & Automation Magazine, 25(1):8–15.
  • Magnanimo et al. (2016) Magnanimo, V., Walther, S., Tecchia, L., Natale, C. and Guhl, T. (2016). Safeguarding a mobile manipulator using dynamic safety fields. V Intelligent Robots and Systems (IROS), 2016 IEEE/RSJ International Conference on, str. 2972–2977. IEEE.
  • Magrini et al. (2015) Magrini, E., Flacco, F. and De Luca, A. (2015). Control of generalized contact motion and force in physical human-robot interaction. V Robotics and Automation (ICRA), 2015 IEEE International Conference on, str. 2298–2304. IEEE.
  • Nguyen et al. (2018) Nguyen, D. H. P., Hoffmann, M., Roncone, A., Pattacini, U. and Metta, G. (March 5–8, 2018). Compact real-time avoidance on a humanoid robot for human-robot interaction. V Human-Robot Interaction, HRI ’18: 2018 ACM/IEEE International Conference on, str. 9. IEEE.
  • Roncone et al. (2016) Roncone, A., Hoffmann, M., Pattacini, U., Fadiga, L. and Metta, G. (2016). Peripersonal space and margin of safety around the body: learning visuo-tactile associations in a humanoid robot with artificial skin. PloS one, 11(10):e0163713.
  • Roncone et al. (2015) Roncone, A., Hoffmann, M., Pattacini, U. and Metta, G. (2015). Learning peripersonal space representation through artificial skin for avoidance and reaching with whole body surface. V Intelligent Robots and Systems (IROS), 2015 IEEE/RSJ International Conference on, str. 3366–3373. IEEE.