The Worldwide Journal of Robotics Analysis, Forward of Print.
This work presents a novel method for the optimization of dynamic methods on finite-dimensional Lie teams. We rephrase dynamic methods as so-called neural extraordinary differential equations (neural ODEs), and formulate the optimization downside on Lie teams. A gradient descent optimization algorithm is offered to sort out the optimization numerically. Our algorithm is scalable, and relevant to any finite-dimensional Lie group, together with matrix Lie teams. By representing the system on the Lie algebra stage, we cut back the computational value of the gradient computation. In an intensive instance, optimum potential power shaping for management of a inflexible physique is handled. The optimum management downside is phrased as an optimization of a neural ODE on the Lie group SE(3), and the controller is iteratively optimized. The ultimate controller is validated on a state-regulation process.
发布者：Yannik P. Wotte，转转请注明出处：https://robotalks.cn/optimal-potential-shaping-on-se3-via-neural-ordinary-differential-equations-on-lie-groups/
