search

Ruben-MR / Quadrotor_3D_planning_control

Repository for the Planning and Decision Making Project Code
MIT License
4 stars 1 forks source link

MPC sample code does not work well, maybe require non-linear optimization #6

Closed xinjie-liu closed 2 years ago

xinjie-liu commented 2 years ago

As the title says, the current sample code only works for a vertical motion (e.g. from [0, 0, 0] to [0, 0, 0.5]). I changed the goal point and some constraints of acceleration, roll angle... it did not work at all. I guess the linearized model in the sample is too weak, maybe we need to implement non-linear optimization in the MPC. But only my personal guess, people who start working on MPC part can take a look at this issue.

xinjie-liu commented 2 years ago

And yes, i linear model cannot work here, the TA of RO47001 Desong Du told me. So we may need to use non-linear one.

xinjie-liu commented 2 years ago

I tried to change the linear model constraints to our non-linear model, but it is hard (impossible, since before optimization process, those values do not exist actually) to get an array from cvxpy.Variable before solving optimization problem. So our focus should be on how to incorporate our non-linear model (equation of motion) to the optimization problem.

xinjie-liu commented 2 years ago

Fixing issue with non-linear MPC