DeepRacer 2023 Racing League
DeepRacer is a unique console experience in AWS as it is not a tool, but a way to generate interest in Artificial Intelligence and Machine Learning (AI/ML) services.
Translating a machine learning process into an easily understood and fun experience is the key function of DeepRacer. The service is a tool to help users to understand basic AI/ML principles and encourage them to dive deeper into other product offerings as a result of using the service.
DeepRacer accomplishes this goal by allowing users to train models that can race a car around a track. Users create, train, refine, and implement reinforcement learning models.
Product
Background
Retaining users is accomplished by continually pushing them to refine their models. The main way this is accomplished is through competition. Much like real auto racing, DeepRacer has a racing league in which users can compete with other users. This is called the DeepRacer Virtual League.
Racing League competition consists of a season that runs from March through October each year. The world championship takes place in November at re:Invent using physical cars on a track.
For 2023, the DeepRacer team had several issues based on user interviews that we wanted to resolve by revising the league structure and user interface. The issues were:
-
All racers compete with everyone in the world which intimidates new racers
-
Racers compete from around the world and have difficulty booking last minute travel to re:Invent
-
Racers who were outside the top 100 in the world had no way to win prizes or gain any recognition
-
Racers who qualified for the finals in October had no reason to race in monthly qualifier races throughout the season
-
Racers with more resources were able to submit models thousands of times to place higher
The 2023 Virtual League should allow racers to compete at a national, regional, and worldwide level with prizes and recognition at each level. In addition, racers should be able to qualify throughout the season for the world championship but should also be encouraged to compete each month.
The Problem
HIGH LEVEL TIMELINE
Four months to conduct research, develop requirements, design league structure, create and revise mockups, complete UX review process, and launch the feature.
THE TEAM
The team consisted of an engineering lead, a PM, a technical writer, and myself as the UX/UI designer.
.
KEY GOAL
To launch a league that would address the known and targeted user issues.
My Role
UX design for the AWS DeepRacer console.
My responsibilities on this project were to work with a researcher to create a survey for racers to help determine league specific issues and prioritize them.
I worked with the PM to write a requirements document based on the prioritized issue list. My function was not only that of a UX designer, but also as a "subject matter expert" as I participate in DeepRacer as well as virtual racing, and also have extensive personal knowledge of auto racing. This allowed me to offer suggestions based on that perspective to how the league structure could more closely resemble traditional racing leagues.
Additional responsibilities were to create and revise user flows and design structure of the league to address the product requirements. I created mockups at varying fidelities and continuously revised the design.
I worked with an engineer to determine what would be possible both within the Polaris(CloudScape) design framework within our timeline.
My responsibilities also included scheduling and completing all UX design reviews prior to launch.
Understanding The User
The target users for this project were both new and existing racers.
Most racers are either students or professionals that work as AI/ML engineers. Many professionals compete as part of a corporate team.
Racers use DeepRacer to not only gain experience, but also to gain recognition to help them get a job in AI/ML, or to advance their career if they already work in AI/ML.
Defining The Solution
The solution for this project consisted of several changes to the structure of the league, which made changes to the user interface necessary.
Below is a diagram illustrating the basic league structure.
The first major change was that we would offer the chance for racers to compete with racers at the national and regional level and earn prizes at each level. This required that we know which country each racer was competing in.
Race entry screens and leaderboards were redesigned to allow users to see their standings in their own country or region, as well as any other county or region.
Below is an example of the race entry screen as well as a national leaderboard.
The second change was the introduction of seasonal points championships at the national and regional level. Each month a qualifying race is held. When a racer submits a time, it is categorized within their nation as well as their region. At the end of each month, a racer is given points towards the national season championship based on their position within their country. They are also given points toward the regional season championship based on their position within their region. The first place finisher in each region for each month earns a spot at reInvent automatically.
Giving racers the opportunity to compete at national and regional levels, allowed for the introduction of season championships. These points based championships encourage all racers, even those who have already qualified, to compete each month for both their national and regional season championships.
This solution, allows racers to know earlier if they will need to travel to reInvent and encourages participation from all racers for the entire season.
Below are high fidelity images of the national and regional season championship leaderboards.
A third change was made to address the number of submissions that a racer could make to each monthly race. To level the playing field, a submission limit was introduced for each race.
While this required some minor UX changes, the more important factor is that we were able to create a simple solution to a problem our users had voiced and implement it as part of this project.
Learnings
Minor changes introduce major challenges
Sometimes changes that seem to be minimal can introduce complications that may not have been easily predicted. By adding granular levels of racing to our league, the amount of information that needed to be presented to the user increased exponentially. This resulted in making decisions about how much to show and when. I prioritized information and used that to determine where it was shown, how it was displayed, and how it would be accessed.
Work with a visual designer (if you have one)
I had not previously worked with a visual designer and this project gave me the opportunity to leverage the knowledge of one. While I determined the hierarchy of information, I relied on our visual designer to help me with font sizes and spacing to present the abundance of information in an easy to read format. This was very valuable on this project as the amount of information to present to the user continued to increase.