For the final design of our Pipe Rider robot surrogate (aka Subway Surfers), it is composed of four sub-systems which include the linear slider mechanism, fixture, camera system, and omni wheels. The combination of these sub-systems is what allows the robot to be manipulated through PVC pipe, turn in various types of PVC pipe, allow the robot to roll 90 degrees, and provide camera feedback footage.
Hardware
Here is the model for the current design of the four-bar fixture. This is what will be mainly used to manually manipulate the robot through the PVC pipes.
Here is the demonstration of how we use the four-bar fixture to turn in an elbow PVC pipe and a tee PVC pipe. Through various iterations and designs, we were able to successfully achieve this turn.
One of the requirements of the robot is rolling 90° along the pipe wall during operation. Unless the user provides extra force to press the wheels against the pipe walls, the wheels can come off of the pipe wall. The linear slider mechanism is responsible for providing this extra force, allowing the user to fully focus on data collection and more complicated fixture manipulation. It is important to note that the linear slider is actuated by 2 total springs.
Another requirement of the robot is to provide camera feedback to the user while being able to rotate the camera 90 degrees. Not only that, but to be able to easily swap the cameras with different models.
Therefore through this design, we are able to successfully rotate the camera 90 degrees and swap it easily through the dovetail design with just one screw
Software
We also use a motorized fixture to keep the robot centered while traversing through the pipes. We use the data from the time-of-flight sensors on the main robot base to determine the error for the motorized fixture to constantly try to center the robot while pushing through a PVC pipe.
As a result of implenting the controls algorithm, one can see the effective of the motorized fixture as it is successfully centering the robot whenever the time-of-flight sensors are off center.
Performance
Encoders are connected to each wheel to record the wheel rotation as it travels through the pipe. In order to accurately simulate the original PipeRider robot and its encoder readings, the wheels must maintain contact with the pipe as it changes orientation. Theoretical simulations are performed in three test scenarios: traveling past a tee, turning into an elbow, and turning into a tee. In each simulation, the robot is assumed to remain oriented vertically and the wheels remain on the same horizontal location of the pipe.
Test in Straight Pipe past a Tee Junction
As the robot is pushed straight past a tee junction pipe, the left and right wheels are simulated to travel 179 and 180 cm.
Comparison of the left and right encoder counts shows that in all three trials, the right encoder traveled a further distance than the left encoder.
Test for Turning in a Elbow Pipe
As the robot robot turns through an elbow pipe. Based on simulation, the left wheel of the robot is expected to travel 185 cm, while the right wheel is expected to travel 171 cm.
The left encoder traveled an average of 975 encoder counts more than the right encoder when the robot turned into an elbow pipe. In both trials, the left encoder measured over 900 counts more than the right encoder.
Test for Turning in a Tee Junction
When the robot yaws 90 degrees to turn into a tee pipe, the left and right wheels are simulated to travel 190 cm and 169 cm.
When the robot was turned into a tee pipe, the left encoder measured 1312 counts more than the right encoder.
Comparison of Final Results
The image on the left is the averaged wheel distances of each encoder for every test scenario. Comparing the difference between the actual and experimental encoder data results in a maximum error of 7.7 cm when turning. Possible causes of this difference may be due to the omni-wheel slipping against the pipewall from loss of contact, or the omni-wheel translating laterally, causing the robot path to be different.