Core Concepts
ASAP-MPC seamlessly combines trajectory tracking and generation, addressing computational delays in online motion planning.
Abstract
This paper introduces ASAP-MPC, a Nonlinear Model Predictive Control scheme for motion planning. It focuses on handling computational delays in real-time implementations by combining optimal control and feedback control. The framework is validated through experiments on drone navigation and truck-trailer manoeuvring applications. Key insights include the methodology, experimental setups, validation metrics, and comparison with the Real-Time Iteration scheme.
Introduction
Challenges in online motion planning.
Importance of trajectory planning for complex systems.
Existing Techniques
Graph-search methods, reinforcement learning, optimization-based planning.
Challenges in real-time implementation due to computational costs.
NMPC Strategies
Fast solvers and specific formulations.
Premature solver interruptions.
Accounting for computational delay.
Methodology
Overview of FUR-MPC, LUR-MPC, and ASAP-MPC schemes.
Illustration of trajectory jumping phenomenon and solution by ASAP-MPC.
Experimental Validation
Applications: Drone navigation and Truck-Trailer AMR parking manoeuvre.
Extensions to previous works and experiment setups.
Validation Metrics
Feedback error analysis for position tracking accuracy.
Comparison with RTI Scheme
Implementation of RTI as benchmark for simplified truck-trailer application.
Results and Discussion
Validation results for drone and truck-trailer applications.
Conclusion
Success of ASAP-MPC in handling computational delays in motion planning.
Stats
NMPC assumes that the solution to the OCP is instantaneously available.
The average computation time using Fatrop is 45 ms for the drone application.
Quotes
"ASAP-MPC seamlessly merges trajectories, providing a smooth overall trajectory."
"The framework successfully deals with computational delays while ensuring feasibility."