/xbot2_interface

Next-gen robot and model interfaces for XBot2

Primary LanguageC++Apache License 2.0Apache-2.0

xbot2_interface

This package replaces the good ole' XBotInterface package by HHCM. We try to keep the API as close as possible to the old one. This is not always possible since this package supports non-Euclidean joints, and therefore the number of joints is in general different than the number of element in the q vector, which is also different than the number of elements in the v vector.

Quick guide

Build the library

After installing the required dependencies (TBD)

mkdir build && cd build
cmake ..
make
make install

Use the library in your project

find_package(xbot2_interface REQUIRED)
target_link_libraries(mytarget xbot2_interface::xbot2_interface)

Model object construction

std::string urdf;  // read urdf file into this string somehow
auto model = ModelInterface::getModel(urdf, "pin");

More constructors are available which take the SRDF file, too.

Vector dimensions

As we support non-Euclidean joints, care must be taken when manipulating configurations and motion vectors.

// the neutral (i.e., zero) configuration 
// note: this is in general different than Eigen::VectorXd::Zero(model->getNq()) !!
Eigen::VectorXd qn = model->getNeutralQ();

// its size is model->getNq()
assert(qn.size() == model->getNq());

// a random configuration complying with joint limits
Eigen::VectorXd qrand = model->generateRandomQ();

// the velocity vector that brings qn to qrand in unit time
// note: you cannot compute the vector difference of qn and qrand,
// since these are NOT vectors in the linear-algebraic sense!
Eigen::VectorXd vrand = model->difference(qrand, qn);

// its size is model->getNv()
assert(vrand.size() == model->getNv());

// the configuration obtained by applying a velocity vector to 
// a given initial configuration
// note: this will be the same configuration as qrand
// note: you can sum two motions, but you cannot sum a motion with
// a configuration
Eigen::VectorXd q1 = model->sum(qn, vrand);  

// the list of joint names
auto jnames = model->getJointNames();

// its length is model->getJointNum()
assert(jnames.size() == model->getJointNum());

// the i-th joint can have nq and/or nv greater than one;
// its index inside a configuration or motion is therefore != id
int id = jnames.size()/2;

// this data structure contains indexing information
auto jinfo = model->getJointInfo(i);

std::cout <<  "id = " << jinfo.id <<
             " iq = " << jinfo.iq <<
             " iv = " << jinfo.iv <<
             " nq = " << jinfo.nq <<
             " nv = " << jinfo.nv << std::endl;
             
// this is how you index a configuration or motion
auto qj = qrand.segment(jinfo.iq, jinfo.nq);
auto vj = vrand.segment(jinfo.iv, jinfo.nv);

// you can use the joint interface (same result);
Joint::Ptr joint = model->getJoint(i);
auto qj_1 = joint->getJointPosition(); // same as qj
auto vj_1 = joint->getJointVelocity(); // same as vj

// similar for effort, acceleration, ...

Using the model to perform computations

// every time you change the model state, 
// call update() to apply the new configuration
model->setJointPosition(qrand);
model->setJointVelocity(vrand);
model->update();

// forward kinematics
Eigen::Affine3d T_1 = model->getPose("my_link");
Eigen::Affine3d T_12 = model->getPose("my_link", "my_base_link");
Eigen::Vector6d v_1 = model->getVelocityTwist("my_link");
Eigen::Vector6d v_12 = model->getRelativeVelocityTwist("my_link", "my_base_link");

// similar for acceleration...

// jacobians
Eigen::MatrixXd J;

if(!model->getJacobian("my_link", J))
{
    std::cerr << "error: bad link name \n";
}

Examples

Check this small collection of commented examples

Porting from XBotInterface v1

Check the cheatsheet