Autonomous Racing — Avoiding going down rabbit holes
When designing any cool systems or products, one struggle is avoiding going down multiple rabbit holes or over-engineering a product without validation. Eric Ries addresses this issue in his book “The Lean Startup: How Today’s Entrepreneurs Use Continuous Innovation to Create Radically Successful Businesses”, where he promotes the idea of the minimum viable product (MVP). Going into details about MVP and incorporating MVP into businesses is out of the scope of this post, but interested readers can refer to the following website http://theleanstartup.com/principles.
This blog series will apply the MVP principles to break down the fundamental concepts behind AI race cars. The application of MVP will help create constraints around the project. Some of the constraints that will be applied to the project in this blog series derive from Nvidia’s DIY AI Race site (https://developer.nvidia.com/embedded/diy-ai-race). The following are the constraints:
- Chassis will be 1/10 RC Scale.
- Restrict the weight to less than 10 lbs (4.5 kg)
- Rely on affordable COTS parts and possibly some custom-built modules but no heavy-duty commercial guidance systems.
- Fully electric (after all, the future is electric vehicles )
- Open-Source software for both running on the car and setting up our cloud environment.
- Video-based driving system
I will refrain from too many constraints since we need to have some room to pivot, but the above 5 are a good starting point. Additional issues that are going to hit this project are costs and the supply chain. At the time of writing, there are shortages in the semiconductor sector and various stock inventories, and the supply chains are all jammed up. As a result, prices for parts are rising, so we will see where cost savings can occur.
For those following along, there is a Git repo where I am starting to store the assets to build an open-AIRacer.