System Performance

Performance Evaluation

Overall, the system had all the necessary components to meet the requirements laid out in the ShipBot specifications. The localization, path planning and mobility platform all had the fidelity needed to successfully navigate the testbed and position ourselves at each station. The initial testing showed that the ShipBot could align with each station with relatively good precision. After more fine-tuning, the ShipBot even could align with different devices without the aid of computer vision feedback. The arm system and end effector could actuate every device in the testbed even when presented with imperfect positioning (could turn valves/flip breakers even if not perfectly centered). In some cases, when there were two or three centimeters of deviation, the grippers still could utilize their high friction foam to grab on target devices and finish required operations. The biggest shortcoming was integrating all systems together. For example, there is still more work that needs to be done to integrate the arm movement and also integrate the camera system. Though the arm movement could perform some basic tasks, the consistency still had much to be desired. The arm movement was responsive, but sometimes it was excessively aggressive, which would shift the entire ShipBot and muddle the odometry readings. If more time was permitted to work on integration (estimated around one more week), the team likely would have a fully-functioning ShipBot.

Strengths and Weaknesses in Design

Strengths

1. Simple control of the arm with linear actuators that required a single PWM signal.

2. Two gripper disks setup reduced complexity and are easy to manufacture and redesign.

3. High friction foam provides enough friction to grab on devices, and have high tolerance for inaccurate localization.

4. Mecanum wheels allow for movement in any direction, and are already heavily used in other robotic projects that the team referenced throughout development.

5. The Arduino Mega 2560 allowed for the integration of many more peripherals in case
   of design changes. Similarly, the use of the Jetson Nano could have allowed us to explore the use of CNN’s for object detection and classification.

6. The use of LIDAR eliminated the need for other sensors such as ultrasonic, and potentially even the camera if navigation is well tuned and tested.

7. The two LiPo batteries can provide up to 30 mins of operation time, even with the stepper motor for the gripper enabled, and other ROS communications between the Jetson and Arduino.

8. The localization program has relatively good accuracy after fine tunings; it could reach designated stations with little deviations.

Weaknesses

1. Arm structure is very stiff and is easily susceptible to the reactionary forces due to valve actuation or misalignment with different parts of the test bed

2. Vertical arm assembly is not sufficiently rigid to withstand torsional load on the z-axis.

3. The current computer vision software for angle detection on valves is not robust enough to function with extreme lighting conditions.

4. Gripper foam might be easily detached from the gripper disks after repeated use.

5. The LIDAR method for localization proves to be less useful when the robot is in a dynamic environment where objects close to the ground change positions; our localization relies on a pre-generated map for the robot to reach its destination properly

6. The ShipBot might not reach its waypoints correctly, causing it to circle around without knowing where it is

7. Arm operation movement might be too aggressive, resulting in inaccurate operation or damaging the ShipBot itself.