基于实时迭代的数值鲁棒NMPC双模稳定预测模型附Matlab代码.zip

# Joint angle control of the 7 DOF robot manipulator KUKA LBR iiwa 14  ## Description This example demonstrates how to use ParNMPC to control a 7 DOF robot manipulator using user-defined dynamics and derivatives. The robot manipulator's dynamics and derivatives are computed using Pinocchio (https://stack-of-tasks.github.io/pinocchio/), which is a very efficient C++ library for rigid multi-body dynamics computations. Since Pinocchio currently only supports Mac and Linux, we here demonstrate how to do the simulation on Ubuntu 18.04. The task is to control the manipulator to its desired reference under the constraints of the input torques and velocities. - Input constraint: |tau| < 10 N - Velocity constraint: |qdot| < pi/2 rad/s - Initial position: [0,0,0,0,0,0,0] - Setpoint in 0 to 4 s: [0,pi/2,0,pi/2,0,pi/2,0] - Setpoint in 4 to 8 s: [pi/2,0,pi/2,0,pi/2,0,pi/2] - Zero gravity ## How to generate code? 1. Open `NMPC_Problem_Formulation.m` and run. 2. Open `Simu_Matlab_Codegen.m` and run. The closed-loop simulation C++ files will be generated to `./codegen/lib/Simu_Matlab` and `./codegen/lib/Simu_Matlab/examples`. The Pinocchio interface is in `./iiwa_pinocchio` (very simple). The urdf file of the manipulator is in `./iiwa_description` (from MATLAB). ## How to run the generated C++ code? 1. Install Pinocchio on Ubuntu (https://stack-of-tasks.github.io/pinocchio/). **Make sure that you can run successfully the two C++ examples in https://gepettoweb.laas.fr/doc/stack-of-tasks/pinocchio/master/doxygen-html/index.html.** 2. Compile and run (supposing the current directory is RobotManipulator). Type the following commands in terminal sequentially: ``` $ mkdir build $ cd build $ cmake .. $ make $ cd ../bin $ ./iiwa14_NMPC ``` ## Computation time In our setup (T = 1 s, N = 18, DOP = 6, i7-8950HK@2.9 GHz), the computation time is about 170 us/iteration. A sampling rate of 1 KHz can be achieved in real time. ## Contributors - Thank [Xuanchen Zhang](https://github.com/xuhuairuogu) (USTC) for providing the wonderful build interface. ## Path following (coming soon) Similarly, you can use ParNMPC to control the end effector to follow a straight line or a sine path ([reference](https://ieeexplore.ieee.org/abstract/document/7575680)):