MPC-MPNet: Model-Predictive Motion Planning Networks for Fast, Near-Optimal Planning under Kinodynamic Constraints

Abstract

Kinodynamic Motion Planning (KMP) is to find a robot motion subject to concurrent kinematics and dynamics constraints. To date, quite a few methods solve KMP problems and those that exist struggle to find near-optimal solutions and exhibit high computational complexity as the planning space dimensionality increases. To address these challenges, we present a scalable, imitation learning-based, Model-Predictive Motion Planning Networks framework that quickly finds near-optimal path solutions with worst-case theoretical guarantees under kinodynamic constraints for practical underactuated systems. Our framework introduces two algorithms built on a neural generator, discriminator, and a parallelizable Model Predictive Controller (MPC). The generator outputs various informed states towards the given target, and the discriminator selects the best possible subset from them for the extension. The MPC locally connects the selected informed states while satisfying the given constraints leading to feasible, near-optimal solutions. We evaluate our algorithms on a range of cluttered, kinodynamically constrained, and underactuated planning problems with results indicating significant improvements in computation times, path qualities, and success rates over existing methods.

A Vision-Based Motion Control Framework for Water Quality Monitoring Using an Unmanned Aerial Vehicle

“…Moreover, recent multi-disciplinary engineering advances in structural design, sensors, actuators and communications enhance the overall autonomy of these platforms. Nowadays, UAVs are equipped with fast and dependable communication technologies [3], enhanced AI-based sensing capabilities [4] and robust motion control schemes (e.g., dynamic window approaches [5], Machine Learning/MPC approaches [6]). In fact, the number of commercially available "ready-to-fly" UAVs demonstrates the maturity of this technology.…”

Chance-Constrained Motion Planning using Modeled Distance-to-Collision Functions

“…In the past few decades, a profusion of work has focused on the motion planning problem for an assortment of tasks such as car navigation around obstacles [1]- [3], constrained robotic manipulation [4], [5], and surgical robot automation [6]. However, most motion planning research has focused on demonstrating examples where environments are highly structured, and uncertainties in sensing are overlooked.…”