SaRA

The tool for Safe Human-Robot Coexistence and Collaboration through Reachability Analysis (SaRA) implements our formally safe approach to reachability analysis in form of reachable sets of humans. The human's reachable sets are given as a union of individual capsules (cylinder with a half sphere at both ends) for each body part. We can formally guarantee safety by executing a failsafe trajectory if an intersection between the human's and robot's reachable sets is detected. We additionally provide classes for simplified linder based reachable sets. SaRA consists of a library (ReachLib) containing all functionalities required for reachability analysis as well as a ROS package (reachable_occupancy) which includes ReachLib and provides volume calculation and visualization in RViz. Both the C++ library as well as the ROS-package can be easily integrated and expanded upon. We utilize live captured Cartesian positions of human joints calculated from motion captured markers. Our efficient implementation is capable of calculating reachable sets and determining intersections between human and robotic reachable sets in only a few microseconds.

SaRA provides the following features through ReachLib:

• Compatibility with Linux and Windows based systems (makefiles are included and can be customized)
• Generation of full body articulated reachable sets and cylindrical reachable sets
• Measurement uncertainties and system delays are accounted for
• Intersection checks of all occupancy containers among themselves and with point clouds
• Fast C++ based implementation
• No additional libraries required
• Extensible hierarchical class structure

SaRA provides the following ROS based functionality through reachable_occupancy:

• Visualization of reachable sets calculated within ReachLib
• Volume calculations for all models contained in ReachLib
• Reception and processing of live motion captured data over UDP and ROS topics
• Compatibility with ros-bags, live and recorded motion capture data, and ROS melodic and kinetic
• Built on fast C++ based nodes
• Integrable in other ROS projects