Academy of Mathematics and Systems Science, CAS
Colloquia & Seminars

Speaker:	Long Cheng
Inviter:
Title:	Recurrent neural networks for convex optimization problems with applications in robotics
Time & Venue:	2022.05.24 10:30-11:10 腾讯会议ID: 740-602-218
Abstract:	Many engineering problems can be converted into certain optimization problems. How to effectively solve optimization problems becomes a research challenge. Since the seminal work of Prof. John J. Hopfield, recurrent neural networks have been applied in dealing with various optimization problems due to their parallel computation nature. In this talk, a recurrent neural network model is presented, which is capable of solving the non-smooth convex optimization problem with the non-smooth convex inequality constraints, the affine equality constraint and the convex admissible range. The feasibility and stability of this recurrent neural network are theoretically analyzed. Finally, two robotic applications are provided: the coordination control of dual redundant robots and the optimal formation of multi-robot systems. It is shown that these two problems can be converted into the corresponding optimization problems and therefore be solved by the recurrent neural network successfully.