複雑な動的環境における収束保証付き自律ナビゲーション
Autonomous Navigation with Convergence Guarantees in Complex Dynamic Environments
閉形式の動作計画とMPCを組み合わせ、複雑な動的2D環境での障害物回避と収束を解析的に保証する制御手法を提案し、経路追従制御へ拡張した。
著者: Albin Dahlin, Yiannis Karayiannidis
分類: cs.RO
原文アブストラクト
This article addresses the obstacle avoidance problem for setpoint stabilization and path-following tasks in complex dynamic 2D environments that go beyond conventional scenes with isolated convex obstacles. A combined motion planner and controller is proposed for setpoint stabilization that integrates the favorable convergence characteristics of closed-form motion planning techniques with the intuitive representation of system constraints through Model Predictive Control (MPC). The method is analytically proven to accomplish collision avoidance and convergence under certain conditions, and it is extended to path-following control. Various simulation scenarios using a non-holonomic unicycle robot are provided to showcase the efficacy of the control scheme and its improved convergence results compared to standard path-following MPC approaches with obstacle avoidance.
