The silent operation was quite a suprising thing. A little shame you had some of the quality issues, but it looks like they could be easily fixed. Overall, I could see so many uses for this little arm. I do love the way things hold together with magnets. That's really great :)
How does this robot compare to industrial robots? Industrial robots are very big and no one makes tiny robots that don't take up much space. This fills that space BUT it looks very slow and quality issues should not happen from a proper robot manufacturer. Is there a way to speed the robot up?
I had recently built a robot with same kinematics, but I am not able to figure out the kinematics part. Is there a way where I can control such type of robot by giving the target coordinates of the end-effector like G-code?
You'd have to do calculations on the fly then submit the joint position via python functions. Im not aware of existing library for this as it takes all params of the robot in consideration
You probably have figured this out by now but i'll give a "quick" summary of what that requires. First G-Code can be used for coordinate based control, but on most printers that is the same as driving the motors directly. We then talk about a Cartesian Work-space and a Cartesian "Joint-space" or "configuration-space". Where the differences between motors steps and the xyz coordinates are constant as they are linearlly proportional to eachother. Y= a*X + b, type formulas. Where a is the scalar and b helps determine the origin. Now to answer your question. Yes you can control this robot giving end-effector coordinates. But the kinematics give you the formula of calculating end-effector location (pose) from the joint configuration of the arm. You are interested in getting the Joint configuration from the end-effector location (pose). One method to do this is called Inverse-Kinematics, subsequently there is a method called Inverse-Dynamics, and both use something called a Jacobian. This is a matrix which is NOT constant, it is a function of the joint-space and it translates the coordinates from work-space to tiny little steps that proportionally move the joints to get the desired end-effector pose. There is a analytic Jacobian, which can be found through differentiation of the kinematics wrt. the joints (often uses Denavit-Hartenberg to derive the forward kinematics) and a geometric Jacobian which is constructed directly from the kinematics using fancy proper mathematics relating to rigid body dynamics and manifolds. In the end the fancy math way is actually easier and faaaar more precise. That said, these methods are not beginner friendly you will easily get a well paying job if you are proficient in this stuff. So the practical solution, use a framework like ROS and let the toolkits developed by thousands of researchers solve the problem for you. If you really want to do it yourself you can simplify the problem to find a solution that works with your current skill set. The first robot arm i buillt had similar kinematics to this, 4DOF. The way i solved it at the time was to remove the motion from the base joint rotating around the Z axis. You are left with an arm that performs planar motion with 2 joints. You can find the trigonometric solution for the forward kinematics of a 2D arm, and with enough effort even invert that equation to get the inverse kinematics as triginometric functions. (no matrices needed) The base joint kinematics is just sin() and cos() functions. When you add it together you can draw lines in 3d space and let it calculate the solution for the joints to the waypoints along these lines. This can all be done with trigonometry Again, this is alot of work, and if you just want a working arm (as is the case with this arm) get into ROS, it is the best practical solution and allows for the most further development of applications and analysis of your arm. The answer is long and might not be relevant to you anymore. But hopefully this is readable, and provides just enough jargon to help people who are still wondering how to get into robotics find relevant information to get the job done online. I highly recommend ROS, their community is great and github not even just for usable code but for reference material to help you develop your own and better programming habits. You can get very far learning only python for analysis and mathematics and C++ for embedded devices such as the Microcontrollers that control the real systems.
