deng-haoyang / parnmpc Goto Github PK

View Code? Open in Web Editor NEW

286.0 19.0 78.0 26.42 MB

A Parallel Optimization Toolkit for Nonlinear Model Predictive Control (NMPC)

Home Page: https://deng-haoyang.github.io/ParNMPC/

License: BSD 2-Clause "Simplified" License

MATLAB 89.72% M 2.06% C++ 2.43% C 4.14% CMake 1.65%

nmpc mpc nonlinear-model-predictive-control model-predictive-control parallel code-generation

parnmpc's Introduction

ParNMPC Version 1903-1

New Features in Version 1903-1:

Primal-dual interior-point method
Improved user interface
Better performance
Line search

Introduction

Homepage: https://deng-haoyang.github.io/ParNMPC/

ParNMPC is a MATLAB real-time optimization toolkit for nonlinear model predictive control (NMPC). The purpose of ParNMPC is to provide an easy-to-use environment for NMPC problem formulation, closed-loop simulation, and deployment.

With ParNMPC, you can define your own NMPC problem in a very easy way and ParNMPC will automatically generate self-contained C/C++ code for single- or multi-core CPUs.

ParNMPC is very fast even with only one core (the computation time is usually in the range of $\mu$s), and a high speedup can be achieved when parallel computing is enabled.

Highlights

Symbolic problem representation
Automatic parallel C/C++ code generation with OpenMP
Fast rate of convergence (up to be superlinear)
Highly parallelizable (capable of using at most N cores, N is the # of discretization steps)
High speedup ratio
MATLAB & Simulink

Installation

Clone or download ParNMPC.
Extract the downloaded file.

Requirements

MATLAB 2016a or later
MATLAB Coder
MATLAB Optimization Toolbox
MATLAB Parallel Computing Toolbox
MATLAB Symbolic Math Toolbox
Simulink Coder
C/C++ compiler supporting parallel code generation

Getting Started (MATLAB 2018b)

Select the Microsoft Visual C++ 2017 (C) compiler:

>> mex -setup

Navigate to the Quadrotor/ folder.

>> cd  Quadrotor/

Open NMPC_Problem_Formulation.m and run.
Open Simu_Simulink_Setup.m and run.
Open Simu_Simulink.slx and run.

Citing ParNMPC

Citing the parallel algorithm:

@article{deng2019parallel,
  title={A parallel Newton-type method for nonlinear model predictive control},
  author={Deng, Haoyang and Ohtsuka, Toshiyuki},
  journal={Automatica},
  volume={109},
  pages={108560},
  year={2019}}

Citing the toolbox (conference version):

@inproceedings{deng2018parallel,
  title={A parallel code generation toolkit for nonlinear model predictive control},
  author={Deng, Haoyang and Ohtsuka, Toshiyuki},
  booktitle={Proceedings of the 57th {IEEE} {C}onference on {D}ecision and {C}ontrol},
  pages={4920--4926},
  year={2018},
  address={Miami, USA}}

License

ParNMPC is distributed under the BSD 2-Clause "Simplified" License.

parnmpc's People

Contributors

Stargazers

Watchers

Forkers

wuyou33 lnshhongbin mr-d-self-driving xuhuairuogu knockin65 yuta527 garlinplus cyy1991 1508189250 deepakingole giegloop lahneh dtbinh minho3 lovesuner harveywen alejandroastudillo yellowpointzd qmqmyu zhou-ust benzpeng saunak626 zzdbirdsfly gaorunzebit heurobo alexe15 shenhm516 ddd-1991 wang-zheng-jiu mayataka fagan2888 ky-lin22 450794083 yinzhenxiao ychepth vinbo pandorathanator xiaozhanggui ztk735 lousample mcheng-zhang linqingbh botvgki changzhengwu chrisjyp layerthinker lucafei hajun0219 yangtao1003 wenbojlu davidzhangbuaa roimar90 congzz niezifei moonieann personqianduixue personal-fork-robot hsxu2mimo ohtsukalab shenggyj gw96 lijian8 3-body solabankole xuannadi liangskywalker roboconer chinahuangyong taroabe mdwsbb01 david-xar mr-guoer walexng wangfj161 projectgp1e3

parnmpc's Issues

Supporting for inequality constraint

Version 1806-1 doesn't support inequality constraints explicitly. However, the method of barrier function requires careful tuning and are not user-friendly.

Functions to deal with inequalities will be included in a future release.

Line search

TODO

Active-set method support

Interior-point method is good for large-scale problems, however, it is not easy to warm start and its barrier parameter is usually set as a constant, and thus only sub-optimal solution is be obtained.

The future version of ParNMPC will support active-set method to deal with constraints.

Unix-style vs Windows-style paths

When running the Quadrotor example, I get following error:

??? OCP_F_Fu_Fx.c specified in custom source files string does not
exist in any of the following search directories:
"/Users/robin/ParNMPC/Quadrotor/.\codegen\lib\OCP_F_Fu_Fx"
"/Users/robin/ParNMPC/Quadrotor"

Code generation failed: View Error Report
Error using codegen

Error in OptimalControlProblem/codeGen (line 123)
    codegen -config cfg_OCP_KKTs...

Error in NMPC_Problem_Formulation (line 53)
    OCP.codeGen();

Error in run (line 91)
    evalin('caller', strcat(script, ';'));

Works if I replace the .\codegen\... path with codegen/lib/....

Failed to load custom code simulation

Dear Deng,
I have tried the Quadcopte example, but the simulink file at the last step didnt run.
I t gives me this error:

Failed to load custom code simulation
library:D:\Users\sanim\Documents\MATLAB\MPC\ParNMPC\Quadrotor\slprj_slcc\oSDpra6UYTzTcubJVwU2TH\oSDpra6UYTzTcubJVwU2TH_cclib.dll
Component:Simulink | Category:Model error

Gauss-Newton Method

Current version of ParNMPC requires the second derivatives of the dynamics, which is impratical for systems with complex dynamics.

TODO: Parallel Gauss-Newton method

Object-oriented programming (OOP) for ParNMPC

OOP based problem definition, code generation, simulation, and deployment are now under development.

confiction with "ss" function provided by the control system toolbox

When I use the function "ss" to create linear state space model after directory ParNMPC with the subfolders included into matlab path, the following error arised:
错误使用 ss
如果类定义了超类，则所有超类必须均为句柄类，或者所有超类均不为句柄类。
In English, which means when the class defined with superclass, then all of the superclasses must be handle class ,or none of them is handle class.

After excluded ParNMPC, the error disappeared.

So, I think the function "ss" maybe conflict with ParNMPC. I have searched the ParNMPC for the keyword "ss", but no result found.
@ideaDeng would you like to find which lines introudce this confliction and fix it.

Failed to run manipulator example

Hi developer,
I got the following error after running the script Simu_Matlab.m

>> Simu_Matlab
Error using f_fu_fx_Wrapper (line 7)
Output argument "dx" (and possibly others) not assigned a value in the execution with
"f_Wrapper" function.

Error in OCP_GEN_fdt_fudt_fxdt (line 2)
   [f,fu,fx] = f_fu_fx_Wrapper(u,x,p,parIdx);

Error in OCP_F_Fu_Fx (line 57)
                [fdt,fudt,fxdt] = OCP_GEN_fdt_fudt_fxdt(u,x,p,parIdx);

Error in coarse_update_func (line 58)
            [F_j_i,Fu_j_i,Fx_j_i] = OCP_F_Fu_Fx(u_j_i,x_j_i,p_j_i,discretizationMethod,isMEnabled,i);

Error in NMPC_Solve_SearchDirection (line 64)
    parfor (i=1:DoP,numThreads)

Error in NMPC_Solve (line 76)
            NMPC_Solve_SearchDirection(x0,pSplit,rho,lambdaSplit,muSplit,uSplit,xSplit,zSplit,LAMBDASplit);

Error in Simu_Matlab (line 73)
    [solution,output] = NMPC_Solve(x0Measured,p,options);

The matlab version I am using is 2021b. And I did not find any other script that can generate the figure in the README file.

macOS Compilation

Hi Deng,

I am using ParNMPC on MATLAB 2018a and macOS High Sierra 10.13.6. While code generation I am getting an error.

`1 error generated.
gmake: *** [NMPC_Iter.o] Error 1

Error(s) encountered while building "NMPC_Iter":

Failed to generate all binary outputs.

??? Build error: C compiler produced errors. See the Build Log for further details.

Code generation failed: View Error Report
Error using codegen

Error in NMPC_Iter_CodeGen (line 83)
codegen -config cfg ...

Error in Simu_Simulink_Setup (line 31)
NMPC_Iter_CodeGen('dll','C',DoP);
`
How to generate dylib insted of dll?

Thank you.

Error NMPC Problem Formulation

Hello,
First, thank you for share with us your work.
I was trying to run the first example but I got the following error

NMPC_Problem_Formulation
Error using sym/vertcat (line 17)
Error using maplemex
Error, (in ArrayTools:-Concatenate) number of columns must match

Error in DynamicSystem/setf (line 5)
obj.f = symfun(f,[obj.u;obj.x;obj.p]);

Error in OptimalControlProblem/setf (line 2)
setf@DynamicSystem(OCP,f);

Error in NMPC_Problem_Formulation (line 33)
OCP.setf(f);

I hope you can help with this.
Best regards,
Julio

Update CMakeLists.txt for pinocchio

Now the CMake packaging for pinocchio is updated and improved.
I think it is nice to fix CMakeLists.txt in RobotManipulator example.
I've done it in my branch, so I'd like to merge the change into this master branch if @deng-haoyang allows it.