Robotic Arm

Bachelor's Thesis.
Portable, with 5-DoF and enhanced maneuverability to handle versatile objects, featuring a human-mimetic end-effector for advanced research in Mechatronics.

Affiliated Company: ENPO SPELS | Feb 2022 - Jul 2022 | Moscow, Russia.
Mentors:

  • Pavel Nekrasov, Associate Professor, Division of Nanotechnologies in Electronics, MEPHI
  • Evgeniy Voloshin, Robotics Team Lead, ENPO SPELS

Thesis [Russian] Presentation [Russian]

Situation:

The objective was to conduct in-depth research in Mechatronics through hands-on full-stack robot development.
Plus, the robotic arm, built by the company’s Robotics Team leader to participate in the RTC Cup Robotics Competition, had big dimensions and inertia as well as limitations to pick most objects but spherical such as tennis balls.

Task:

  • Revise the entire robot’s mechanical design to refine and optimize for functionality and portability. Build it using FDM 3D printing.
  • Design custom PCBs to meet the updated robot requirements, and manufacture them using photolithography technology.
  • Develop software to harness the robot’s full potential.

Action:

With the technical guidance and support of mentors and SPELS, the following were accomplished:

1. Mechanical Desgin & Manufacturing:

To address the issue with grasping spherical, axially symmetric, or irregular-shaped objects I followed biomimcry - human fingers. Actually three pairs of them:

But for simply grasping without complex maneuvers, having three or six servos to control every finger would result in a heavy end-effector.
So, for grasping/ungrasping I decided to rotate all three pairs of fingers simultaneously. But when any of the pairs and/or fingers touches the object’s surface first, it should stop rotating still applying some force to hold the object without ceasing the other pairs/fingers from rotation.
One of the solution was to use drums shown below, and rubber bands to transfer torque as ligaments do in human body.

The mechanism shown below on the left consists of two separate gears with saw transmission (3, 4), spring to push the gears to each other (2), and fixtures (1, 5).
The upper gears of each pair of fingers is rotated by a single motor, while the lower gears rotate each finger individually.
When any of the fingers stops rotating due to contact with the object surface, torque from the upper gear would be much less to rotate the lower gear.
As a result, the upper gear starts jumping up-down without stopping the other gears, still applying torque to fix its finger’s orientation.
The design for a single pair of fingers shown below on the right was the first attempt.
It didn’t work because of rubber extension and slippage between gears as a result of its elacticity, that I looked for to pass it through gear mechanism.
Double-edged sword.

To address the rubber issues above, I switched to drums and capron thread/fishing line (of several diameters) shown below.
The choice was given to capron thread rather than fishing line because of its elacticity and resistance to sun light and low humidity:

As the drums with fishing line could apply greater torque than gears with rubber band, small motors for drones couldn’t provie the corresponding torques.
So, I switched to a more powerful motor and changed the wrist design correspondingly:

Once the wrist started to grasp objects, then next step was to design the rest of the arm.
The first version shown below faced significant issues such as huge inertia because of a peripheral servo at the elbow joint.
Disclaimer: Almost every commercial robotic manipulator has its servos at its joints, which I think is the best approach as long as high motion precision at stake. And I came to this conclusion after I tried with different transmission mentioned in this project below.

So, I transferred the servo from the elbow joint to the shoulder joint, and transferred torque through rubber band.
But it lead to the same outcome - extension because of its elacticity, which significantly affected the end-effector pose repeatability (error in reaching the target pose in space given it’s the same at each time).

Apart from switching from rubber band to capron thread, the 5th degree of freedom was introduced to increase manipulability.
But this version was far from the optimal due to its some parts with redundant mass which didn’t have any significance in the overall durability.

Finally, after rigorous material science analysis to make the design durable and lightweight, the model shown below was built:

Once the mechanical part of the robotic arm was finished, additionaly its base and the remote controller were designed and printed:

Mechanical design of the Robotic Arm base (left) and the Remote Controller (right).

2. PCB Design/Manufacturing:

Electric Components
Structural diagram of the Remote Controller (left) and the Robotic Arm (right).
Schematics of the Remote Controller (left) and the Robotic Arm (right).
PCB layout of the Remote Controller: top (left) and bottom (right) layers with/out ground fill.
PCB layout of the Robotic Arm: top (left) and bottom (right) layers with/out ground fill.

Photolithography requires positive/negative masks with the PCB layout during the exposure stage.
Additionally, there should be an external frame to align the masks before the exposure.

Outer frame for visual alignment in the development stage of photolithography (left) and the overal PCB layout w/o ground fill (right).
Top (left) and bottom (right) layers were converted from positive (2nd row) to negative (3rd row) masks due to Ordyl Alpha 350 specifications.
Final masks should be black-white due to Anycubic Photon UV integrated light (wavelenght 405nm).

Photolithography process consisted of:

  • Substrate (Copper Textolite) Pre-treatment.
  • Photoresist Coating: 110℃ heated rollers.
  • Exposure: 405nm UV light; 110sec.
  • Development: 1% water solution of Na2CO3; ultrasonic bath; 30-32℃; 24.4mins.
  • Hard Bake: 150℃; 2mins.
  • Etching: hydrogen peroxide(150g) + lemon acid(30g) + NaCl(5g); 30-32℃; 50mins.
  • Via Drill/Connection: mechanical.
  • Additional Metallization/Isolation.
Poor quality photoresist (not Ordyl Alpha 350) resulted in bad results during the development stage.
The entire process of photolithography and post-process treatment.
The final PCBs of the Remote Controller (top) and the Robotic Arm (bottom). The Remote Controller's PCB was the first trial.

3. Software Development:

  • Development Tool: STM32CubeIDE.
  • Development Language: C/C++.
Program flow for the Remote Controller. The entire program (left) and the control input reading (right) from buttons/joysticks.
Program flow for the Robotic Arm.

Results:

  • DoF: 5
  • Mass (kg): 1.5
  • Dimensions (mm): 300 x 150 x 200
  • Payload (g): 170
  • End-effector: Gripper
  • Control: Remote
  • Input Voltage (V): 7-40
  • Programming Language: C/C++
  • Cost (₽): 5000