This week CROSS is beginning to prepare for Prototype Inspection Day (PID), by identifying potential demos for November 3rd. We maintain consistent communication with our liaison engineer for their expectations of our progress and our demo during PID.

This week the power supply ordered for our LiDAR base sensor has been delayed, thus effecting our potential of integrating the LiDAR into our prototype at this point in time. In the meantime, the team remains focused on the development of training an object recognition LiDAR model to recognize choice objects that we can reasonably demo for PID.

Additionally, our team has continued evaluating potential camera sensors for later camera fusion data. The main focus of camera sensor evaluation this week is the IR cameras that KB Signaling currently owns, that being the ZN1A-DNT352XE-M and ZN1A-B4DZF69U, are considered for temporary use as they are security cameras that are more optimal for static detection rather than dynamic detection.

Going into next week, CROSS will be focusing on refining our plan of action for our PID demo and our prototype presentation. Additionally, the evaluation of potential IR camera sensors will be expanded to include camera models used to create the IR and thermal IR dataset used in autonomous vehicle object detection research literature.

This week was one of the most exciting weeks for Team CROSS as we went to Melbourne, Florida to visit KB Signaling. The trip was roughly 3 hours from Gainesville, so our team spent a lot of the morning together on the road to get to their facility. It was particularly interesting to visit Melbourne and see the environment in which our sponsor operates, as there is a large engineering and technology community in Melbourne.

Our tour of KB Signaling’s engineering lab was the largest portion of the trip. We had an opportunity to observe some of the systems and products that the company works with, such as older technologies that have been applied to train localization and railway signaling. The systems we saw in person helped us relate the research we have been conducting in the semester to the actual equipment and engineering challenges in the railway industry. It also provided us with a deeper insight into how our localization project using LiDAR would be able to coexist with the current railway technologies.

A large portion of the visit was dedicated to our sponsors for our Preliminary Design Review (PDR). We shared our research results, proposed system design, current progress, and direction we intend to take going forward. It was a successful presentation and a wonderful chance to share our design thinking process with the people who will be judging and utilizing the work we create.

On the visit, our liaison Tony gave us a bit of a lesson in singing and shared some of his experience of producing music. It was a great break from the routine of engineering and project talks and a good opportunity to get to know our sponsor a little better.

In general, the visit to the site was very helpful in providing Team CROSS with a clearer understanding of the actual context of our project. The trip was useful in linking our semester of research with the actual engineering problem we are trying to solve, as well as in seeing the engineering lab and the railway technology first hand, and successfully presenting our PDR.

In this past week, Team CROSS were preparing for our site visit to KB Signaling at Melbourne FL. Specifically, we completed the visiting logistics, including car rental and trip planning. Furthermore, we incorporated the feedbacks, finalized our PDR presentation and is ready to present to our Liaison. 

We also made good advancement on the literature reviews. The two route we found in our research to approach the train localization problem include LiDAR only method and a camera IMU fusion method. Both method share a common theme of finding landmarks along a trip and fuse that information with odometry to locate the train. LiDAR only method relies on point clouds to detect landmarks and utilizes LiDAR odometry. Whereas camera IMU fusion method would rely on vision models such as YOLO to detect km post or mile post, which that information will be fused or be used to correct the IMU inertial outputs. 

Lastly, we also downloaded OSDaR23. This dataset, collected from onboard, contains 2D/3D annotations for RGB, IR Camera, LiDAR, and Radar frames, serving as the standard evaluation dataset for uni and multi-modal system in rail setting. The image displays are the RGB, IR camera, LiDAR points viewed in BEV and 3D of a randomly sampled frame from OSDaR23.

This past week was another busy one for Team CROSS as we wrapped up our Preliminary Design Review (PDR) and started getting more hands-on with the Aeva Aeries II LiDAR sensor.

A big part of the week was spent presenting our PDR and getting feedback on our project scope, system architecture, technical performance measures, and overall project direction. The feedback gave us a better idea of what we need to clarify and improve as we continue developing the system.

On the technical side, we finally started working directly with the Aeva Aeries II and AevaViz. We spent time getting the sensor connected, working through power and networking setup, and trying to visualize live LiDAR point-cloud data. There were definitely some troubleshooting challenges along the way, but the process helped us learn a lot more about how the sensor and its software actually work.

Now that the PDR is behind us, we can shift more of our focus toward collecting LiDAR data, building out our point-cloud processing pipeline, and eventually using that data for railway landmark detection and classification.

We’re also excited for our site visit with KB Signaling next week! We’re looking forward to seeing more of their work firsthand, learning more about the environment our system is being designed for, and spending some time with our sponsors. We’ll be sure to keep everyone updated and share plenty of photos from the visit!

Overall, this week felt like a good transition from mostly planning and design work into more hands-on testing and development.

This past week was a busy one for Team CROSS as we continued making progress on both project planning and technical development.

We drafted our project roadmap, created a Jira project management board, developed a project risk table, and continued working on our preliminary system architecture. We also reviewed Aeva Aeries II documentation and discussed how the sensing and software components will work together.

On the technical side, we began training a YOLO object detection model using railroad-related imagery as an early step toward landmark recognition and as a baseline we can continue improving with more representative data.

A large part of the week was also spent preparing for our Preliminary Design Review (PDR) next Tuesday, including drafting and practicing our presentation, refining our proposed design, and organizing our project plan and technical direction.

Over the weekend, we also had the opportunity to participate in the White Belt Quality Workshop presented by Florida Power & Light. It was interesting to learn how Lean Six Sigma principles can be used to improve and optimize engineering processes. We’re grateful to Linda Linnus and Dr. Doreen N. Ayafor for sharing their experience with us.

Overall, it was a productive week of planning, meetings, technical work, and PDR preparation as we continue moving toward the next stage of the project.