I have a basic understanding of inverse kinematics, in that there's a workspace and a configuration space, and the point is to find a solution in configuration space that satisfies a workspace goal. Multiple solutions may exist in configuration space for redundant manipulators.
That said, I'm currently struggling with a problem where the workspace is a plane with blocks (cubes) placed on it, and I need to plan a path for moving a block in the workspace while avoiding collisions with other blocks. Importantly, the arm (manipulator) moving the block must not collide with other blocks as well. The latter situation could be possible in case blocks are stacked to form a tower.
I recognize that planning a path for block to avoid other obstacles is a simpler problem that may be solved with A* and the likes. However, I'm interested in solutions that also account for obstacle avoidance for the manipulator. For my part, I have looked at the existing research (mainly robotics journals), where Rapidly Exploring Random Trees (RRTs) and their variations seem to be discussed a lot. Before I spend too much time learning and implementing those approaches, I'd like to confirm if I am on the right path, or are there any approximate solutions to this with low runtime cost that I should look at? Thanks.
Note: I've made some omissions from the question in order to keep it simple, but I'd be happy to answer any questions asking clarification.