Modelling of Rotor drag dependency on quadrotor velocity and thrust

$latex \dot{p} = v$ $latex \dot{R} = R\hat{\omega}$ $latex m\bold{\dot{v}} = -mg\bold{z_{W}}+TR\bold{e_3}+R\bold{F_a}$ $latex I\dot{\omega} = -\omega \times I\omega +\tau_{g} + \tau + \tau_{a}$ $latex F_{a} = \sum_{i=1}^{n}F_{a}^{i} = \sum_{i=1}^{n}-\sqrt{T_i}c_{d1}\pi_{e3}V_{a}^{i}$ $latex V_{a}^{i} = R^{T}v + \omega \times d_{i}, \pi_{e3} = I - e_{3}e_{3}^{T}, T_i = k_T \omega_i^{2}$ $latex \sqrt{T_i}$ Current models $latex T_i = T_h = … Sigue leyendo Modelling of Rotor drag dependency on quadrotor velocity and thrust

Practical 6DoF object pose and size estimation

YOLO $latex L = \lambda_{coord}L_{coord} + \lambda_{coord}L_{size} + \lambda_{obj}L_{conf} + \lambda_{noobj}L_{\bar{conf}} + \lambda_{class}L_{class} $ Grabner, A. et.al $latex L = L_{proj} + \alpha L_{dim} + \beta L_{reg}$   Singleshot6dpose $latex L = \lambda_{coord}L_{coord} + \lambda_{obj}L_{conf} + \lambda_{noobj}L_{\bar{conf}} + \lambda_{class}L_{class}$ $latex L = \lambda_{coord} \sum_{i} [(x_i - \hat{x_i})^2 + (y_i - \hat{y_i})^2] +  \lambda_{conf}\sum_{i}(C_i - \hat{C_i})^2 … Sigue leyendo Practical 6DoF object pose and size estimation

Installing and using singleshot6DPose

Hello! So this time I will be installing and trying to use singleshot6DPose. This is an state-of-the-art neural network for object pose detection using RGB images. It is based on YOLO so its superfast. I am starting with a system as follows: Ubuntu 16.04LTS Cuda compilation tools, release 9.2, V9.2.148 Python 3.5.2 Nvidia GeForce 1050 … Sigue leyendo Installing and using singleshot6DPose

3D Quadrotor: LQR Controller for the non linear model (using differential-flatness property)

1) Proof of differential flatness for 3D quadrotor 2) LQR Controller 3) In the previous post, I derive the proof of differential flatness for the basic model of a quadrotor. With the equations developed in that post, it is possible to compute trajectories for all the states and inputs so that a quadrotor moves from … Sigue leyendo 3D Quadrotor: LQR Controller for the non linear model (using differential-flatness property)