Khái niệm cốt lõi
Analyzing the solvability of the Inverse Optimal Control problem based on the minimum principle.
Tóm tắt
The content discusses the solvability of the Inverse Optimal Control (IOC) problem using two existing minimum principal methods. It aims to determine which trajectories of the original optimal control problem will result in recovering the true weights of the reward function. The analysis focuses on open-loop and closed-loop systems, comparing soft and hard-constrained methods. Various mathematical conditions are provided to verify if a trajectory contains sufficient information for recovery. The paper also validates analytical results through simulations and provides insights into systematic ways to find optimal control weights.
Thống kê
Initial research by Kalman for IOC in linear systems [4].
Machine Learning community's work on MDP setting [5]-[8].
Proposal by Mombaur for humanoid robot motion replication [9].
Formulation by Keshavarz and Johnson based on necessary conditions of optimality [10], [1].
Works based on minimum principle method [2], [11], [12].
Molloy's utilization of Riccati solution for soft-constrained IOC method [2].