# Toolfitting in Path planning (manipulator)

• posted

Dear all, I am currently working on a 4 degree of freedom robot that has to move in cartesian motion, thus requiring me to firgure out its inverse kinematics. Being a 4DOF robot working it cannot achieve arbitrary positions and orientations but this is not what I care about. I want the robot to only achieve proper positioning. The problem I am facing is that my invese kinematics force the robot to maintain orientation and not position (I am guessing since the joint angles are primarily derived via the 3X3 rotation part of the forward kinematic matix). Is there a way to estimate an orientation while making the position of the tool the primary goal to be achieved. I have heard of a process called "Toolfitting" but so far haven't found anything on the topic via google or looking up the univesity catalogue. Regards and thanks for the help.

• posted

It would help if you explained the configuration of your robot's axes. Are they revolute or prismatic? [aka rotating or linear?] Do you have Denavit-Hartenberg parameters or the like? What do the kinematics look like?

All things are possible until details are added.

- Daniel

• posted

Thank you for your reply. I realized after I posted that there might such a question asked. Unfortunately I am not allowed to divulge this information but rather I am wondering if there are any general purpose method that allow estimation of the approach vector. I can say that I am trying to find four joint angles using only three parameters (the position vector), which I believe is not possible. While I can't give out information about the geometry of the robot (which suuuucks) I was wondering what might the best way(if any) to estimate an approach vector. I can say that the 1st three joints keep the robot in a plane while the motion of the last two move the robot out of the plane.

• posted

Well, all the approaches I know start by looking for symmetries and decomposing the problem into smaller pieces. For example, with a two-link revolute arm, most solutions have an "elbow-up" and an "elbow-down" variant; the law of cosines can be used to solve whichever one you want.

Given what you've specified, I might try decomposing the problem into a "primary-plane" part and an "out-of-plane" part. First find solutions of the out-of-plane joints to give you the desired height z. Then try solving the in-plane joints to get the required x-y coordinates for each of the out-of-plane solutions. Finally use whichever combination seems best.