Lie MPC for AMVs

Convex Geometric Trajectory Tracking using Lie Algebraic MPC for Autonomous Marine Vehicles

IEEE Robotics & Automation Letters

paper link: https://ieeexplore.ieee.org/document/10301632

We propose a computationally-efficient Convex geometric error-state MPC for marine vehicles. The proposed algorithm is compared with Nonlinear MPC.

Used packages

1) OSQP : QP solver

Folder : osqp

Installation is required to run the code.

2) CasADi : NMPC solver

Folder : casadi-linux-matlabR2014b-v3.5.5

3) MPC-tools for CasADi

Folder : octave-mpctools

Tool for easy implementation of NMPC with CasADi by autonomously transforming the given dynamics model into the proper format.

4) USV Otter model from MSS

Folder : GNC, HYDRO

Control input of the otter model is changed from motor speed to thrust force.

Demonstration

Run main.m in MATLAB

Parameters

• Reference trajectory, ocean currents, prediction horizon, simulation parameters are defined in main.m
• MPC cost weights (P, Q, R) are defined in sim.m
• For calculating dynamics of USV Otter, NMPC uses otter.m and NMPC-simple uses otter-simple.m. They regard ocean current as 0 m/s. True simulation is processed using otter_true.m with ocean currents.

Results

Trajectory tracking results

Computation time (msec) for single optimization

Ocean Current	Proposed MPC	NMPC	NMPC-simple
0 m/s	49	764	478
0.5 m/s	50	953	460