Some planners assume a kinematic model of the robot for example diff drive or omni directional etc. while generating motion plans for the robot. Can knowledge of the dynamics of the robot help generate better motion plans ? If yes, why is this not typically done apart from the difficulty in identifying or coming up with such models ?
Yes it helps but it definitely adds complexity to the planners. Depending on the wheel configuration, how many wheels, the mass of the robot, the moments of inertia, same as the mass of the wheels, moment of inertia of the wheels, and friction as well. All this parameters add more complexity that is hard to generalize. For this reason all these planners assume a robot that follow the specified velocities to a certain accuracy - the low level controllers take care of this part, while the planner acts as a high level controller.
As an extra you can take a look at this video of Disney that improves the trajectory tracking of its animatronics by knowing their dynamic model. https://www.youtube.com/watch?v=Z1jgaEO9aRs