🏎️ Why you need DeepRacer 🏁
Meet our CIOs
In the vibrant halls of South Hills High School, Chief Information Officers Denise and Sophie embarked on a mission to infuse the DeepRacer Autonomous Race with a sense of identity and purpose. Recognizing the importance of a memorable acronym to encapsulate the essence of the project, they dove headfirst into the creative process, determined to craft something that would capture the imagination of all who encountered it.
Drawing inspiration from the innovative spirit and dedication of the South Hills community, Denise and Sophie set out to create an acronym that would reflect the heart and soul of the DeepRacer race. As they brainstormed ideas and explored different concepts, they kept coming back to one central theme—the pursuit of excellence, precision, and perfection.
It was then that the acronym SHARP—South Hills Autonomous Racing Program—was born. Symbolizing the dedication to innovation and the relentless pursuit of autonomous racing perfection, SHARP captured the essence of the project in a succinct and impactful way.
But creating the acronym was just the beginning. Denise and Sophie knew that for SHARP to truly resonate with the community, it needed to be more than just a collection of letters—it needed to be a rallying cry, a call to action, and a symbol of unity and pride.
With SHARP in hand, Denise and Sophie set to work bringing their vision to life. They designed eye-catching banners and flyers adorned with the SHARP logo, ensuring that every corner of the school was imbued with the spirit of the DeepRacer race. They also collaborated with the business department to create stylish shirts emblazoned with the SHARP acronym, allowing participants and spectators alike to wear their pride on their sleeves.
But perhaps the greatest impact of SHARP lay in its ability to inspire and motivate. As Denise and Sophie shared the acronym with their peers and the wider community, it became a symbol of possibility and potential—a reminder that with dedication, determination, and a little bit of creativity, anything was possible.
As race day dawned and the excitement reached a fever pitch, SHARP stood as a beacon of hope and inspiration, guiding participants and spectators alike on a journey of discovery and innovation. And as the cars roared to life and the cheers of the crowd filled the air, it was clear that SHARP was more than just an acronym—it was a testament to the power of collaboration, creativity, and the boundless potential of the human spirit.
Sophie has strong technical and creative skills! 
Denise was the 1st SHARP Champion! 
The SHARP banner hung outside the CS Lab all year! 
Both Sophie and Denise were very talented at Machine Learning!
Meet our Engineer 1s
Amidst the hum of excitement surrounding the DeepRacer Autonomous Race, Engineer 1s Maddie, Armando, and their team stood as unsung heroes, their contributions often overshadowed by the technical prowess of their peers. Yet, despite lacking the advanced technical skills of their counterparts, Maddie and Armando proved to be invaluable assets to the project, their roles serving as the glue that held the entire endeavor together.
Maddie, with her boundless energy and can-do attitude, took charge of the logistical aspects of the project, ensuring that every detail was meticulously attended to. Tasked with setting up the boundary walls around the race track, Maddie approached the task with gusto, armed with nothing but a handful of zip ties and a determination to see the project through to completion. While her role may have seemed simple compared to the complexities of machine learning algorithms and microprocessors, Maddie’s attention to detail and unwavering commitment to setting up the track barriers was the foundation of the DeepRacer Event.
Meanwhile, Armando, with his steady demeanor and quiet determination, took on the responsibility of assisting with the physical races. Running alongside the DeepRacer cars with a portion of the boundary wall in hand, Armando played a crucial role in ensuring the safety and smooth operation of the races. Despite the physical demands of his role, Armando approached each task with grace and precision, his unwavering dedication to the project serving as an inspiration to his peers.
While Maddie and Armando’s roles may not have involved complex algorithms or cutting-edge technology, their contributions were no less important to the success of the DeepRacer race. Without their tireless efforts and unwavering commitment to excellence, the project simply could not have been completed.
Meet our Engineer 2s
In the realm of the DeepRacer Autonomous Race, Engineer 2s Sarah, Riley, Bryan and James stood out as the driving force behind the technical prowess and competitive spirit that defined the project. Armed with a deep understanding of Reinforcement Learning and a thirst for competition, they not only set the tone for the race but inspired their peers to push the boundaries of what was possible.
Sarah, with her analytical mind and passion for machine learning, dove headfirst into the world of DeepRacer competitions. Through countless hours of training and experimentation, she honed her skills to perfection, dominating the online Student League races and virtual competitions with ease. Her success on the track served as a rallying cry for her peers, inspiring them to join her in the pursuit of racing excellence.
Meanwhile, Bryan, with his keen eye for detail and strategic thinking, approached the DeepRacer race as a chess match, carefully analyzing each move and optimizing his models for maximum performance. With a thirst for competition rivaled only by his passion for innovation, Bryan pushed himself to the limit, determined to emerge victorious in every race he entered. At one point Bryan took off his shoes and ran in front of the car hoping that the agent would follow his white socks! It worked and his agent won the 1st ever physical race on the classroom track.
And then there was James, whose boundless enthusiasm and infectious energy brought a sense of excitement and camaraderie to the project. In addition to his competitive spirit, James took on the responsibility of maintaining the DeepRacer cars’ batteries and organizing the garage where they were stored—a task that was as crucial as it was often overlooked. Despite the technical complexities of the project, James approached his role with enthusiasm and attention to detail, ensuring that each DeepRacer car was equipped with a fully charged battery before every race.
But perhaps most notably, James found himself in a unique position within the team—he became the student, learning from and being guided by his teammate, Engineer 3 Anthony. Recognizing the value of embracing new opportunities for growth and development, James humbly accepted Anthony’s mentorship with open arms. With patience and generosity, Anthony shared his knowledge and expertise with James, guiding him through the intricacies of the Terminal and bash commands to set up localized training. Through their shared experiences and mutual support, James and Anthony forged a bond that transcended their roles within the project, creating a dynamic partnership built on trust, respect, and a shared commitment to excellence.
Together, Sarah, Riley, Bryan, and James formed a formidable foursome—a powerhouse team that set the tone for the entire DeepRacer race. With their technical knowledge, thirst for competition, and unwavering commitment to excellence, they inspired their peers to push the boundaries of what was possible, driving innovation and fostering a sense of camaraderie that would define the project for years to come. And as race day dawned and the cars roared to life on the track, Sarah, Bryan, Riley and James, and their fellow Amazon Future Engineers stood at the forefront of the action, ready to showcase their skills and prove that in the world of autonomous racing, anything was possible with dedication, determination, and a little bit of competitive spirit.
Meet our Engineer 3s
Anthony, a master of the Terminal and bash commands, assumed a pivotal role in setting up the localized training environment on the department’s PCs. With his meticulous attention to detail and troubleshooting abilities, Anthony meticulously configured the machines, ensuring optimal performance for racers to hone their skills with precision. His expertise became a cornerstone of the project, providing a solid foundation upon which the team could build their technical prowess.
Meanwhile, Joel’s path to the project was a tale of serendipity and curiosity. Despite hailing from a class unrelated to DeepRacer technology, Joel was captivated by the sight of Raspberry Pi’s and DeepRacer cars, drawn to the project by an innate thirst for knowledge and a passion for electrical engineering. His decision to join the team marked the beginning of an incredible journey—one filled with challenges, growth, and camaraderie.
Collaborating closely with Anthony, Joel eagerly embraced the challenge of setting up the finish line for the race—a task that demanded both technical expertise and innovative thinking. Leveraging his background in electrical engineering, Joel approached the task with creativity and resourcefulness, ensuring that the finish line not only functioned flawlessly but also showcased the team’s ingenuity and dedication.
However, as with any ambitious project, challenges inevitably arose. It was during these moments that Anthony’s troubleshooting abilities truly shone. With his quick thinking and methodical approach, Anthony adeptly tackled technical issues as they arose, ensuring that setbacks were swiftly overcome and progress continued unhindered. His ability to think on his feet and find solutions to complex problems became a linchpin of the team’s success.
Together, Anthony and Joel formed a formidable team—a dynamic duo whose complementary skills and unwavering determination propelled the project forward. Their collaborative efforts, resilience, and commitment to excellence served as an inspiration to their peers, fostering a culture of innovation and teamwork within the project.
Now, let’s delve into the challenges that the event presents. The conflicts are numerous, and managing the project often feels like coordinating the many arms of an octopus. However, with the right team members and sufficient time, these issues can be swiftly resolved. Successfully executing this project requires a significant investment of time—far beyond what is typically available during a regular school day.
Many times cars would all of a sudden stop working. Meaning that they would stop connecting to the private network that DeepRacer sets up for the cars. We spent countless hours troubleshooting these cars and at one point tried to ‘flash’ the car with a fresh install of the DeepRacer stack. When dealing with these cars I found that it was quicker to just return the cars to Amazon and get a new one. Amazon was pretty good about it and it expedited the fix. However, while doing research on one of the cars that quit on us I found that the certificate was out of date. Updating the certificate fixed the car. It had me wondering if this was the fix to all the cars that we returned.
We also purchased a new mesh wi-fi system for the race. This mesh network enabled us to put more nodes around the room for which we thought allowed our agent to never lose connectivity. Not sure this is how mesh networks work, however, the router came with a nice app that helped you understand connectivity issues and let you reset the router. Resetting the router was a fix for us a couple of times when we had connectivity issues.
We had a major battery failure on the day of the race. We brought extra batteries, but 3 out of 5 vehicle batteries failed on us on race day, leaving us with only 2. During the year the batteries never failed us, in fact, they have been so solid that we forgot that the batteries came with a ‘starter’ cord to turn them on if you ever found them to be turned off. We never needed the cord, the batteries seemed to always work. I imagine the batteries turned off to prevent them from being overcharged. Since it had been such a long time, if ever, that this had taken place, we forgot about the ‘turn on’ cord and deemed the batteries as failed. It was the DeepRacer Community on Slack (now on Discord) that brought my attention to this fact. When we applied the start cord to the dead batteries, they turned on!
The track barriers purchased through Amazon have a bit of a learning curve to get right. 1st it takes one stand connected to two panels, zip-tied together at the top and the bottom to get the whole thing to work. This takes a lot of trials but we found this to be the best way to attach them.
Once you attach them, you can leave them together except for the four corners that make the rectangle around the track. We unsecured the four corners and had the students daisy chain the barriers to and from the event. You can store the walls around the perimeter of your room, for use for a later race or event.
Mark Ross had a great point that I didn’t read until after our event. Per Mark, make sure that you remove or don’t use locations with unnecessary lights or white fixtures. When we signed up to have our race in the Student Union, I didn’t realize that the school media board would be constantly running and couldn’t be turned off. The light that shined on the track created quite the glare and the car had trouble dealing with the excess white in its camera. The windows of the Student Union also created quite a glare. Next year we will try to do the race in the windowless gym!
The last part of the setup was the finish line. This part was added on at the end and wasn’t part of our initial goal. However, with some encouragement from the Machine Learning Community, specifically Mark Ross, we were able to pull it off. It turns out that this would be an essential part of making our race look pro. It wasn’t that hard, though we went through (2) Raspberry Pi’s, and the look was so much better than calling it from someone’s stopwatch on their phone which is prone to major inaccuracies.
We used this setup since time was limited:
https://github.com/aws-deepracer-community/deepracer-timer
Next year we will try to pull off this one:
https://github.com/aws-solutions-library-samples/guidance-for-aws-deepracer-event-management/
The suggested hardware is on the GitHub page itself. But definitely assign this project to a team member and have it available for your race. If applying for a grant make sure to allow for extra funds to purchase the hardware.
Additionally, I purchased wire to enable long runs and a beginner breadboard starter kit to connect the hardware to the computer.
Raspberry Pi 4 8GB
Scheduling conflicts are certain especially when considering the amount of time it takes to train the students on Machine Learning. We started by getting the students competing in the Student League, then into the Virtual League. We then showed them how to add their models to the physical car and raced them around my classroom track. After showing them localized training, the students took some time to adjust to using the terminal ( it also takes time to get those machines ready for use).
So we scheduled a Spring race date, which was a perfect amount of time to ready every thing and every student for the race. Plus, the school was able to better handle the event on this late date (March 27th).
This late race date also gives you ample time to invite members of the community and other schools and teams to your event. I invited Computer Science teachers from the Dave Wittry Programming Contest who have conducted this memorial programming contest some 15 years and have an exhaustive list of CS teachers in Southern California. The only school that seemed interested is the one where the Wittry Contest hails from, Troy Tech in Fullerton, California. After an email back and forth, I never heard from the program director again and it seems, the autonomous racing discipline just might be too new to spike anyone’s interest. That invite went out to some 80 schools and/or teachers.
It is extremely difficult to get your head around the DeepRacing platform and Machine Learning subject. So spending time educating those around you can be difficult to manage. But to be a promoter of Machine Learning you must get the word out to as many as possible and share your findings as often as possible.
It was troublesome to acquire help from the community, however I was able to get community involvement via our students’ parents. Peter Aspiras and Anthony Torres fathers of James (Engineer 2) and Anthony (Engineer 3) respectively, came to our events and classroom to see what we were working on. This generous help was not looked upon lightly as Peter is an ex-Raytheon Engineer, and Anthony worked with lasers for an unnamed company. To have valued community members take part in our program is an honor and a good start to expanding our reach.
With our 1st race in the books, I will now try to take on the journey of growing the community locally.
Procuring the Trophies would be a challenge if it weren’t for our CIOs Denise and Sophie. A quick search on the Internet and they soon found out what a racing trophy should look like and ordered one up from an online trophy shop. However, listing their story on the project scrum board, Joel (Engineer 3), piped in and let it be known that his aunt owned a trophy company. So the three team members conspired to have the trophies made, “in-house.” Always check around to see what resources are right in your midst because this made the trophies easy to acquire and less expensive also.
Keeping track of a such a large event can cause a lot of anxiety. However, with the use of a Scrum or Kanban board we were able to piece out the project in smaller chunks. We use these boards in our AP CS A class and found out about them through various software development sites and agile systems. Tasks have color coded levels of difficulty. Each task is usually put into categories of type of engineer who could be working on it i.e. Engineer 1 thru 3 or CIO.
As a teacher, I determine what level the task is and place it on the board. Engineers then decide what task that they want to complete and choose to work on those tasks, trying to move the sticky note across the board to done. This visual allowed for us to see how much work needed to be done and if someone was taking too long and needed help.
The notes at the top were the color code of type of engineer who would most likely be successful completing the job. Engineers were told that they could work on any job, and that the color code was just a suggestion on how hard the job might be to complete without the proper experience. I had a scrum board for each part of the race that we were working on. For the DeepRacer event, it took 6 different scrum boards. You could put the entire event into one scrum board but this is how we decided to break it down.
Acquiring the necessary funds to conduct the race is a huge conflict. You will need at minimum $5000 to be able to conduct the race. The majority of the funds necessary come from purchasing an AWS custom track. Purchased from Amazon, the RL Raceway Track cost $1600. Track Wall Barriers were the bulk of the cost coming in at $2400.
In California, I was able to write a grant to procure the necessary funding for the event (See Below). Other items needed were an AWS DeepRacer Car, Trophies, and accessories. The accessories were mainly for the finish line, (Raspberry Pi and sensors). Other considerations that would have made the event pro-level are shirts for the DeepRacer Team and possibly swag for all contestants like stickers, water flasks and keychains.
To find the necessary time to pull off the race, our team met during lunch (30 minutes) everyday. We started the lunch crew with Maddie (Engineer 1) James (Engineer 2) and (Anthony Engineer 3) meeting every school day starting about two month’s in.
Secondly, Engineer 2 James and Engineer 3 Anthony would meet after school 1 day a week starting second semester. This meeting would usually last 2-3 hours on Thursday. This extra time allowed us to troubleshoot technical issues and test the car’s training.
Many of our questions were answered by the AWS Machine Learning Community on Slack. The group has transitioned to the AWS Machine Learning Community on Discord. Without this talented community none of this would be possible. Anyone trying to conduct a race needs to be part of this community.
Our goal was to empower high school students within the Computer Science Department to compete in both the DeepRacer Student League and Virtual League, and to conduct a physical race. We are humbled to have successfully achieved all three, and we deeply appreciate the transformative potential of this endeavor. It provided us with a unique opportunity to foster a new generation of innovators and problem solvers through immersive, hands-on experiences with cutting-edge technology.
As I reflect on the journey we’ve undertaken together, I am filled with a profound sense of gratitude for the opportunity to be part of such a remarkable and inspiring community. Our DeepRacer Autonomous Race was not just an event—it was a testament to the power of collaboration, innovation, and the unwavering belief in the potential of every individual to make a difference in the world. And as we look towards the future, I am filled with hope and excitement for the endless possibilities that lie ahead.