I think that the simplest algorithm you can use is standard RRT.
It is up to you to decide to use it in task space or in joint space.
Be careful that describing the position of simple obstacles (like a box for instance) in the task space is stragthforward, while it might become very difficult if you operate directly in joint space.
In joint space, it is possible to use the trapezoid acceleration curve for space path and trajectory planning. If the desired path is defined in Cartesian space, then the transformation of waypoint coordinates to the joint space is provided by using inverse kinematics. In the case of populated environment (obstacles existence), you can use sampling based algorithms (RRT, RPP, EST) or road map algorithm (PRM, QRM,...).
are you considering a 2D or 3D problem? Do you want to run over the obstacles or consider them as having infinite height? Configuration space is less intuitive in the beginning but it simplifies control equations, I recommend that you get used with both representations.
I agree with Mr. Matteo Regalia. Its upon to you decide in which space do you want to specify a trajectory. What I can add is to opt for some algorithm whose third derivative of position is continuing to avoid generations of jerks (vibrations at high speeds). In the package that was passed to you, the trajectory generator generates straight lines in Cartesian space (or operational space), using trapezoidal speed profile as is normal in industrial robots. But this generator generates jerks since its acceleration curve has discontinuities. It would be better to use a single third or fifth order spline algorithm, but note that 3 or 5 intermediate points between iniciail and end points must be specified to raise the desired polynomial. The inverse kinematics block provided in the package can handle with positions, velocity and accelerations between cartesian and joint space, since, depending the control algorithm you decide to use, its requires the desired joint accelerations.