Compact Robotic Wrist and Gripper Using Stewart Platform Geometry

Deliverables

Requirements

Specifications

Structure and actuator selection

A structural analysis was completed on 4 hexapod configurations (6-6, 3-6, 6-3, 3-3) to determine the most suitable design. Based on its forces and manufacturability, Type 6-6 was chosen. The picture below shows our analysis for one of the configurations.

Based on the results a Pugh Matrix was draw and found that the ideal structure to use was type 6-6 despite not being the one that has the least stress, given that it still met the specifications needed for our project.

Next, the actuators were chosen with a similar process. With the team’s first CAD iteration, different actuators were used on the model to check parameters and determine which one be the best for our application. Below are the three actuators for which the selection was done.

Based on the results the following Pugh Matrix was draw and had the L12-I as the team’s recommended actuator.

A dynamic model and fatigue analysis were created to evaluate operation of the device before construction. The dynamic simulation modeled worst-case forces acting on the platform and actuators at various orientations using Siemens NX. Using harmonic movement, the maximum force in a single actuator reached 2.5 N during nominal movement and 14N with lockup— both within the L12 actuators’ 22 N capacity.

Using the maximum force a single U-Joint pin would experience; the fatigue analysis satisfied the infinite life criterion for both the Soderberg and Modified Goodman methods ensuring no number of cycles would cause failure.

Two of the team members worked on creating the tool paths and generating the G-Code required to run the HAAS 5-axis CNC mill. Each of the team members worked on one specific part, one on the bottom connecting block, and one on the bottom connecting block Sawyer connecting plate. The resulting parts can be seen below respectively.

Another of the team members focused on manufacturing the top and bottom stewart strut plates using the manually programmable ProtoTrack CNC mills. This was because instead manufacturing with this specific CNC would be much faster as unlike the HAAS CNC mill there is not as extensive review process to manufacture the parts, allowing the team to move faster in the manufacturing of the device. A picture of the bottom and top plate can be seen below respectively.

Another team member was responsible for making top connecting pins, manufactured using the manual lathe. An 8mm rod with the correct thread was purchased and the processed on the lathe via different operations until the pins were made.

Another teams member was responsible for 3D printing and assembling the gripper. Apart from the screws, the heat inserts and motor, the entire gripper is 3D printed.

Although the electrical and computer engineering team was responsible for controlling the Stewart platform, a simple control system was developed to collect preliminary test data. This system involved controlling each actuator using an Arduino, linear potentiometer and a power supply. A control code was created to operate the actuators through an interfacing linear potentiometer.

The Quantum X FaroArm® was used to scan the Stewart Strut assembly, measuring displacements relative to the base to verify range of motion and system accuracy. The setup involved calibrating both base and top plates and actuating the platform to its extreme positions to capture maximum translation, tilt, and twist. Testing confirmed that all design specifications were met, including the expected Z-translation limit of three centimeters due to actuator stroke constraints.

Next, the size constrain (diameter) was tested by having a 3D printed 8cm cylinder and checking that the device fitted inside the agreed envelope (diameter only, as height was also checked with the FaroArm). The device weight was also checked using a scale and made sure that it was not over 1.5kg, as the entire device plus payload had to be 2kg at maximum and a minimum payload of 0.5kg was required.

The two videos below show early tests of the gripper working and making sure a tennis ball could be held in place, as well as the platform moving (sped up). The control system from the Electrical and Computer Engineering team is design to move the platform a lot faster. The Mechanical Engineering team’s simple circuit was just used to show proof of a working device and to drive it to testing positions.

In terms of design, the base plate configuration has the largest room for improvement. To begin, the connection between the sawyer base plate and the platform base plate could be simplified and better constrained. Rather than using three screws with extruded knobs, which requires the aluminum knobs to bend inward to fully constrain the plates, a latch connection could be used with only one fastener. This would lessen the material needs of the two plates and better constrain the plates together.

Another improvement to the base plate configuration could be shelling more material out of the center of the platform base plate. There is only a need for material on this plate in a thin outer ring where the dowel pins, PCB screws, and sawyer plate connection lie. Otherwise, material in the center could be completely removed to save weight. Having a gap in the middle could also leave room for the PCB to be sunk into the base plate, rather than sitting on top of the plate. This could allow the dowel pins to be shorter, lessening their chance of bending and failure, and would reduce the total height of the device.

In terms of manufacturing, there could be an improvement on the way that the sawyer plate and bottom plate are manufactured and assembled. A suggestion to improve the process is to machine each of the two parts profile individually first (no side holes to connect them). Then make sure they fit together properly and make a boss to hold both pieces together through the middle hole they both have. Once that is done, they can be machined as an assembly. First on the lathe to get a flush and clean finish and then on the mill alongside an indexing head to drill the side holes all together. This would ensure that the holes are perfectly aligned and at exactly 120 degrees from each other. In the prototype presented on this paper, this was not the case and when assembling, the user must make sure to align the witness holes on the side to make sure both parts fit together with no issues and can be screwed easily.

Moreover, another suggestion would be to make the universal joints from scratch. Making them this could allow the user to machine, say on of the universal joints side and the connecting pin as a single part removing the need of set screws which have proven to be complicated to work with due to size of the device. In a similar way, the press fitted pins, risers and universal joint on the bottom plate could all become a single part making it stronger and easier to work with. When making these part and additional alignment feature could be added so that the actuators position is defined from the start. In the current assembly, this causes some issues as the user must align them and perfect alignment cannot be guaranteed as it is all done by “eye”.