Dynamic Optimization Fabrics for Motion Generation

Optimization fabrics are a geometric approach to real-time local motion generation, where motions designed by the composition of several differential equations that exhibit desired behavior. We generalize this framework dynamic scenarios and nonholonomic robots prove fundamental properties can be conserved. show convergence trajectories avoidance moving obstacles guaranteed using simple construction rules components. In addition, we present first quantitative comparisons between optimization model predictive control generate similar with better scalability, thus, much higher replanning frequency (up 500 Hz 7 degrees freedom robotic arm). Finally, empirical results on robots, including mobile manipulator 10 human, supporting theoretical findings.