Initial Design¶
1 Background¶
Introduction¶
vscout is an open source software suite attempting to simplify and enhance scouting for all robotics teams. This can help teams pick alliance partners, generate strategies against opponents, and know this information in real time. It also is designed to be easy to use and significantly more time efficient for the scout’s organization compared to traditional methods. Customizability is a key factor influencing the design of this software suite, allowing the end user to collect any data they would like.
Features of Final Solution¶
- Cross-Platform
From a Windows, Linux, MacOS, Android, Oak OS, Fuschia OS or iOS device, one should be able to easily input data through their respective GUIs (web and mobile) or CLIs (desktop only).
- Data Administration
The data will be available to those specified by administrators of that particular instance of the data set. This allows some users to only read data, some users to have full access, and some users to have no access.
- Data Analysis
Through machine learning and/or data analysis, a user can choose to let their device compute expected statistics like expected autonomous paths.
- Flexible
There is no limit or minimum requirement to the number of devices used to collect and store data. The software can store any type of quantitative data, and be able to display statistics in the user’s desired way.
- Easy to Use
vscout focuses heavily on UX Design, so if there are any UX issues, it will be a top priority to fix. It uses Material Design by Google, which is an established and researched set of design practices to have natural, easy to use UIs.
- Real Time
Data is available in real-time and can be visually shown in the form of a graph or any type of visual representation appropriate for that type of data.
Problem¶
With regular scouting, there is no established good practices or methods to scout. Oftentimes, beginners are left to scout, since they cannot contribute to the robot or talk about the robot in great depth. More often than not, these beginner scouts will not be able to do a proper job without some training. With vscout, the job of scouting should be effortless, efficient, and data-driven. Extra scouts can gather the same information about the same match, and reduce bias efficiently. For experienced scouts, the ability to gather, and analyze data efficiently is what differentiates vscout from regular scouting.
Non-Goals¶
vscout is not a replacement for scouts, and is a tool for them to use to make their job more efficient. vscout does allow the option for text input for some qualitative data, but its use case benefits much more from quantitative data.
2 Milestones¶
V1¶
Our MVP is a CLI and mobile app that can store and perform basic analysis on many forms of data by itself on one device. There should be the option to have multiple data sets.
V2¶
Version 2 should allow multiple instances of local databases (different devices) to be synced and allow for data administration.
V3¶
This version should help predict various statistics and include more rigorous data analysis. There will also be integration with VEX DB if connected to the Internet to get match statistics automatically.
3 CLI Language¶
Python¶
- Click Library
- Developers already familiar
- Very easy to develop in
- Scripting language
- Consistent SQL database if using SQLite
Dart¶
- Flutter cross-platform UI toolkit
- Args and cli_util libraries
- Will get developers familiar with Dart for Flutter app
- Consistent SQL or NoSQL database if using both SQLite or sembast
- Hot reload
- Smooth and fast
- Easy to learn
Decision¶
Dart¶
Although the click library for Python is amazing for making CLI programs, and Python networking libraries are top notch, the Flutter framework for mobile is amazing. It supports all top mobile platforms including the myserious, upcoming Fuschia OS. With one code base, development time is saved, and with easily customizable Material Design, the UI will be easy to customize. Dart itself is not a bad language and is heavily web-oriented. Its async library is top notch and it has support for a NoSQL and SQL database through these libraries: Sembast, and SQLite. It is possible to write CLI programs in Dart, so the logic from the CLI programs can be transferred directly to the Flutter-based mobile app. If we implement an MVC design pattern, this might aid the transfer of logic.
4 SQL vs NoSQL¶
Sembast and vs SQLite
SQL¶
- Uses SQL
- Very powerful
- Very versatile
- Widely used
- Pre-defined schemas
- All data must follow same structure
- Up front preparation
- Change in structure is disruptive and difficult
- Vertically scalable
- Table-based
- Powerful Queries
- SportsDB super complex structure
NoSQL¶
- Dynamic schema for unstructured data
- Can create documents without having to first define structure
- Each document can have unique structure
- Syntax can vary between databases
- Add fields as you go
- Horizontally scalable
- Data stored in many ways: column, document, graph or KV
- No standard interface to perform complex queries
- Queries themselves are not as powerful
Decision¶
NoSQL¶
SportsDB is very, very, very complicated, and very hard to properly grasp. Implementing it is not too hard since they provide the source on their website. Working with the data from SportsDB would be a huge challenge. The reason it is so complicated is due to the fact that it can be flexible. NoSQL is flexible but less complicated, so it is much better for this use case. Although NoSQL’s queries are not as powerful, this should not pose a problem with proper data structures. With NoSQL databases, the user will have no issue during a new season, since the database will not need updating every season. It also will help when the user customizes what data they want to collect.
5 Distributed vs Centralized¶
Distributed¶
- Master copy on client
- Harder to understand
- Harder to abstract
- Less likely to have permanent loss of data from one failure
- Always be able to show data from own device
- Can be disconnected
- Local actions are fast since it’s local over requesting server
- Changes can be sent whenever
Centralized¶
- Easier to make
- Single master copy on one device (server)
- Access control easier
- Must request centralized server for current dataset
Decision¶
Distributed¶
A centralized system makes the user experience much worse, so the end user may find vscout too complicated, and revert back to regular scouting without the app. Although a distributed system is more complicated to build, it is the far superior solution. One thing to consider is having a distributed system but have one machine have the master copy of the information, so mimicking a centralized system. This can be explored later during the development of V2.