kotakku / simple_casadi_mpc Goto Github PK

hello. I've recently been trying to solve an manipulator motion planning NMPC problem using the pinocchio robotics library. However, I am experiencing problems using the pinocchio library in stage cost and terminal cost function.

I am trying to solve the following motion planning problem using nmpc.

I am trying to solve forward kinematics by taking the q value in the state from the stage cost and terminal cost, but the state is of MX type so I cannot use the pinocchio library. In the case of inequality constraints, the distance between the robot and the obstacle must be measured using the robot's q value for obstacle avoidance, but it is also difficult because the state is of the MX type.

Would it be helpful to change the arguments of the function to the casadi::DM type?

Hello, I am a student trying to implement MPC based on your library. I would like to express my gratitude to you for creating a library that made it easy to use casadi, which made it easy to implement mpc.

The question I would like to ask in the repository is that if the predict horizon is smaller than the number of states, nx, a segment error occurs.

The part where the problem occurs is in the code below that declares the MPC class in simple_casadi_mpc.hpp.

for (auto l = 0; l < nx; l++)
  {
         lbw_.push_back(x_bounds[nx - 1].first[l]);
         ubw_.push_back(x_bounds[nx - 1].second[l]);
  }

I would like to ask whether the number of repetitions in the for statement in the above code should be predict horizon N, not nx.