Goal-Conditioned Terminal Value Estimation for Real-time and Multi-task Model Predictive Control

• DOI: 10.48550/arxiv.2410.04929
• Publication Date: 2024-10-07
• Authors: Miyo Terao Morita; Satoshi Yamamori; Satoshi Yagi; Norikazu Sugimoto; Jun Morimoto
• Keywords: Model Predictive Control; Value (mathematics)

While MPC enables nonlinear feedback control by solving an optimal problem at each timestep, the computational burden tends to be significantly large, making it difficult optimize a policy within period. To address this issue, one possible approach is utilize terminal value learning reduce costs. However, learned cannot used for other tasks in situations where task dynamically changes original setup. In study, we develop framework with goal-conditioned achieve multitask optimization while reducing time. Furthermore, using hierarchical structure that allows upper-level trajectory planner output appropriate trajectories, demonstrate robot model able generate diverse motions. We evaluate proposed method on bipedal inverted pendulum and confirm combining real-time control; thus, successfully tracks target sloped terrain.