(Mechanized Autonomous Rail Scanner)
SpaceX has expressed a need for a rail and sub-track visual observation system by which an operator could identify abnormalities or foreign object debris (FOD) on the track, and we get to build it! However, we've decided to go a few steps further with our designs.
In order to run the scanner system, a robotic transportation system is needed to shuttle our imaging solution down the rail. The Mechanized Autonomous Robotic Scanner or “MARS” is designed to be an especially stable and reliable shuttle on which our imaging systems will be housed.
MARS is designed to complete one scan of the Hyperloop tube within 10-20minutes (MARS can go faster, but visual inspection becomes nearly impossible). In addition to visual inspection, MARS provides SpaceX with the width of the gap between every I-beam pair at a resolution of less than 0.05 inches and a graph indicating FOD as a function of displacement down the tube. This core data is supplemented with relevant statistics we automatically generate.
This summer, a robust software package, nicknamed Goddard, was created to control MARS. This software controls every aspect of MARS, interfacing with systems such as the video streaming setup or our circuits, acquiring and forwarding telemetry to a human operator, and self-monitoring itself for any erroneous values which may indicate a critical issue. If certain thresholds are crossed, then MARS recalls itself automatically to the tube entrance where it can be inspected and repaired. We have also built a graphical interface which will allow anyone to operate and monitor MARS without any major difficulties and without a severe learning curve. An entire scan can be completed with one command.
If you are interested, all of our code can be found here: https://github.com/JoeBartelmo/PyDetect
The electrical system is also highly coupled with Goddard, as any component on the robot can be turned on or off at will. This allows us to have a power-save state in which MARS can hibernate in the tube for hours on end if required. Signals from several different sensors are processed and forwarded along as telemetry. Fail-safes and watchdogs are also built into the hardware itself to recall MARS in the event of any problems.
If you are interested, our circuit designs can be found here: http://bit.ly/2cyWN9m
As far as inspection itself goes, we have 3 "action cameras" on-board which provide a live feed to human operators at a resolution of 1280x960px. These are on adjustable mounts currently, and are able to be adjusted prior to runs in order to maximize the visibility of the rail and subtrack as needed by the operator. A custom image processing package has been implemented in order to make the operator's job easier as well, highlighting regions that have been calculated to contain objects not expected to be in frame for a specified camera view. This is displayed over top of the camera feeds in real time.
In addition, we have two high-speed Ximea sensors which image the rail at roughly 100fps from directly above. This imagery is processed in real-time and an array of distances between points along the gap between I-beams is generated. Our conservative math suggests a resolution smaller than .05 inches, but testing is ongoing for the empirical verification of the system's resolution. In the near future, it is possible this system will be altered to detect braking damage on the top of the rail as well.
All in all, this robot will be invaluable to SpaceX as they evaluate pod teams' designs and test them physically in the coming months. We hope that it will serve well its purpose and that our project and ideas will be implemented and improved upon as Hyperloop becomes a reality!
P.S. We suppose this qualifies for working on a MARS mission with SpaceX, so we plan to pad our resumes accordingly.