Create vehicle motion model - Githubissues

gt-marine-robotics-group / Virtuoso

ROS2 autonomy architecture for Georgia Tech Marine Robotics. Designed to be modular and work with any combination of sensors + motors.

8 stars 3 forks source link

Create vehicle motion model #2

Closed ragingonyx closed 2 years ago

ragingonyx commented 3 years ago

We need to create a mathematical representation of our vehicle dynamics. Currently we have an empirical model, aka our physical vehicles, but this isn't ideal for tuning and testing in simulation. I hope to be able to leverage the empirical model in the creation of the theoretical model in a system like this.

Here's a useful link to get started with: Video explaining how to make models.

ychitale commented 3 years ago

AE8900_REPT_Rameshbabu_Rahul.pdf

ychitale commented 3 years ago

Currently looking into whether the dynamic model from VRX can be extracted and used. This would be ideal because:

It's a high-fidelity model that includes: "a 3D visual model, dynamic and hydrodynamic characteristics, collision model, coefficients to characterize wind effects and non-linear static thruster behavior"
It is customizable to a degree, and tutorials for customization are available: https://github.com/osrf/vrx/wiki/customizing_vrx_tutorials This includes custom setups for the thrusters. In fact, the holonomic configuration is already available as one of 3 standard configurations: https://github.com/osrf/vrx/wiki/tutorials-PropulsionConfiguration
Mass properties for the WAM-V are already defined

ychitale commented 3 years ago

Papers that describe the WAM-V dynamic model used in VRX: Towards_Maritime_Robotic_Simulation.pdf Station_Keeping_Control_USV.pdf

colineRamee commented 3 years ago

FYI: I think it's a great first step but the dynamics of the model are still a bit weird so it might not translate super well to real world (when I was designing the controller I thought the simulated wam-v was like a bar of soap, the real wam-v does not glide that much, it's just my opinion though I don't have data to back that up).

ychitale commented 3 years ago

Got it, thanks! I guess we might have to see if we can tune the model so that it matches the real-world behavior better.

ychitale commented 3 years ago

System_Identification_Using_the_Modeling_through_Iterative_Smoothing_Algorithm.pdf

ychitale commented 3 years ago

https://github.com/osrf/vrx/blob/master/usv_gazebo_plugins/src/usv_gazebo_dynamics_plugin.cc : Hydrodynamic equations for the WAMV in VRX are implementedt here

https://github.com/osrf/vrx/blob/master/wamv_gazebo/urdf/dynamics/wamv_gazebo_dynamics_plugin.xacro: Dynamic parameters are set here

ychitale commented 3 years ago

Building a simplified model based on Fossen's equations of motion (3DOF) as stated in: R. Rameshbabu, M. Steffens and D. Mavris, "System Identification Using the Modeling through Iterative Smoothing Algorithm," Global Oceans 2020: Singapore – U.S. Gulf Coast, 2020, pp. 1-7, doi: 10.1109/IEEECONF38699.2020.9389220.

fossen_eqns_3dof

Initially using values from VRX for hydrodynamic coefficients and mass properties