1. Locomotion Base

We noticed that many teams used omni-directional wheels in their bases for locomotion. As some of our members had experience with omni-directional locomotion bases in the past, we decided that it would be also the best option for us in terms of simplicity, but also flexibility in maneuvering through the test bed.

Full System Sketch of our Robot Design with Omniwheels in our Locomotion Base

2. Turntable Base

Although none of the previous teams had implemented a turntable base in their robot designs because of the omni-wheels that allowed them to rotate in place, we noticed that in tight spaces such as in the corner of the test bed, it would be difficult for our robot to rotate perfectly in place. Thus, we decided that a turntable would be novel way for use to turn in place quickly at the corner of the test bed.

Turntable Sketch

3. X and Z Gantry

Since one of our members had prior experience with the Hebi motors, he felt that instead of having a highly complex robot arm, it would be best to have a 2 degree of freedom gantry that will help position the base of the arm so that the remaining degrees of freedom would be easy to solve for in the inverse kinematics. In addition, we questioned the TAs and learned that the team that incorporated a 2 DoF gantry last year did well because the gantry could exert a strong normal force to the objects and was more precise. Thus, we decided to incorporate a X and Z Gantry on top of our turntable.

X-axis Gantry Sketch

Z-axis Gantry Sketch

4. Robotic Arm

Finally, we decided to go with a simple 3 degree of freedom robot arm similar to the one used by the team with the 2 axis gantry last year because we found that it would be enough to position our end-effector to actuate any object in the test bed. Our arm is can be modeled as a planar revolute revolute (RR) arm, which makes our inverse kinematics extremely simple.

Arm Sketch

5. Granular Jammer Gripper

Finally, we observed some videos of grippers from the past year and decided that a granular jammer gripper would work best as our robot's end-effector and would be simple to implement. In addition, one of our prior members had experience building one previously, which increased our confidence that it would work.

Between our two designs at the start of the project, we chose to go with the granular jammer design. With its ability to conform to any surface that is pressed onto, the granular jammer can actuate objectives with greater ease than a friction pad. The geometry of our overall system allowed us to easily overcome the challenge of installing the vacuum and tubing systems for the device, as the hole sizes of the HEBI motors made it easy to route the tubing to the granular jammer. Furthermore, by having access to several fabrication tools, we were able to overcome the challenge of mounting a camera to the end-effector. This was achieved by making a camera mount that attached to the HEBI motor, a much more rigid structure than the granular jammer itself.

Robot Arm

In order to position the arm and end-effector to accurately come into contact with the objectives, the robot must have a system that both aligns the arm with the objective and brings it within a suitable range for the end-effector to have optimal contact. To that end, our team thought of two systems that accomplish this goal. Our first design consisted of a biaxial gantry configuration that would be capable of translating our robot arm in both the x- and y- directions. Our second design consisted of a 4-DOF arm that would be capable of moving in all 3 linear directions, as well as rotating on the XZ-plane. The table below details the scoring comparison of our two design ideas.

Robot Arm Decision Analysis

Between these designs, we decided to go with the biaxial gantry system. While its positioning capabilities for the end-effector are not as versatile as the 4-DOF arm, it is able to provide motion ranges and forces to a greater degree. Additionally, it is much easier to program and control than the 4 DOF arm. This reduced the positioning error that our end-effector encountered, resulting in more successful operations.

Rotation Mechanism

Given the configuration of the testbed, where some of the objectives are perpendicular to each other, our robot will need to rotate in some fashion in order to actuate all of them. To overcome this obstacle, we devised two designs for rotation. Our first option was to rotate the entire robot by using our locomotion system. With the omni-directional wheels in our base, we would be able to have our entire robot rotate to properly face the objectives that it needs to actuate. Our second design is a turntable system. The use of a turntable system would allow us to rotate our gantry and end-effector subsystems without having to rotate our entire robot. The table below details the scoring comparison of our two design ideas.

Rotation Mechanism Design Analysis

Of these designs, we decided to go with the turntable subsystem as our rotation mechanism. Despite the design being less favorable in terms of cost and fabrication, the system will provide much more precision in properly positioning the end-effector. Plus, while our locomotive subsystem will still have omni wheels installed, using a turntable for rotation will allow us to only move the robot along the X and Y axes individually, which increases the precision of our base localization.

Vision Systems

For our vision system, we ideally wanted to be able to sense depth to help with robot localization and sensing the environment. Thus, we conducted a brief trade study shown in the table below between two popular depth cameras: the Intel RealSense D435 and the Microsoft Kinect v2.

Vision System Analysis

Overall Robot Design Mock-Up