Efficient contact models for realistic human-robot interaction in simulated environments

Robotic and Autonomous Systems (RAS) are a critical component of military operations on the battlefield of the future. RAS are capable of performing complex, dangerous tasks, thereby reducing risk to Service members. Extraction of wounded individuals from dangerous areas so they can receive medical attention is an ideal application of RAS. However, there has been relatively little research and development of RAS for medical applications in the field. One roadblock to developing medical RAS is the lack of realistic digital human models for simulationbased prototyping of design and control of robots that physically interact with humans. Simulation can be an important element in the RAS development workflow, allowing innovation and experimentation in rapid iteration cycles without risk of damaging expensive RAS hardware or risking human injury. This project will address the gap in the simulation toolset by developing anatomically realistic, physics-based digital human models for use in robot simulation environments. Detailed models of the human musculoskeletal system are available in open source tools such as OpenSim, but these models are based on rigid body dynamics and lack soft tissue properties. Therefore, the focus of this effort will be on developing enhanced soft tissue contact models for simulating a robot manipulator grasping, pushing, or palpating various regions on the human body. The soft tissue models will be implemented as spring-damper elements. The stiffness and damping properties of the model will be tuned using a high-fidelity finite element model of the corresponding anatomy. The use of a simplified model tuned to match a high-fidelity model will allow good approximation of the true soft tissue behavior in a computationally efficient implementation. This efficient contact model will be integrated with a full rigid body dynamics model of the human body (e.g., OpenSim). The full integrated model will be capable of simulating contact force at the grasp location, distributed load across the body surface, muscle strains, and forces at the human body joints. This model can be readily integrated into robot simulation environment, as it utilizes components (springs and dampers) commonly available in these environments.