On making robots understand safety: Embedding injury knowledge into control

Sami Haddadin, Simon Haddadin, Augusto Khoury, Tim Rokahr, Sven Parusel, Rainer Burgkart, Antonio Bicchi, Alin Albu-Schäffer

Research output: Contribution to journalArticlepeer-review

151 Scopus citations


Enabling robots to safely interact with humans is an essential goal of robotics research. The developments achieved over recent years in mechanical design and control made it possible to have active cooperation between humans and robots in rather complex situations. For this, safe robot behavior even under worst-case situations is crucial and forms also a basis for higher-level decisional aspects. For quantifying what safe behavior really means, the definition of injury, as well as understanding its general dynamics, are essential. This insight can then be applied to design and control robots such that injury due to robot-human impacts is explicitly taken into account. In this paper we approach the problem from a medical injury analysis point of view in order to formulate the relation between robot mass, velocity, impact geometry and resulting injury qualified in medical terms. We transform these insights into processable representations and propose a motion supervisor that utilizes injury knowledge for generating safe robot motions. The algorithm takes into account the reflected inertia, velocity, and geometry at possible impact locations. The proposed framework forms a basis for generating truly safe velocity bounds that explicitly consider the dynamic properties of the manipulator and human injury.

Original languageEnglish
Pages (from-to)1578-1602
Number of pages25
JournalInternational Journal of Robotics Research
Issue number13
StatePublished - Nov 2012
Externally publishedYes


  • Physical human-robot interaction
  • biomechanics
  • human injury
  • impact dynamics
  • nonlinear control
  • robot standardization
  • safety
  • soft-tissue


Dive into the research topics of 'On making robots understand safety: Embedding injury knowledge into control'. Together they form a unique fingerprint.

Cite this