The hardest challenge is autonomous vehicle development isn’t technical – it’s raising enough money to keep going 💸
The hardest technical challenge, however, is determining when you are ready to remove the safety driver and let the system operate on its own. Once you’re able to make that assessment, the task list to get to “ready” becomes clear, and progress is steady.
Kodiak just published a blog post on two of the major tools we use to determine when we are ready to “go driverless.” We use these tools every day, and we’ve used them to go driverless in an increasing range of domains. We’re using them to prepare to go driverless on the highway.
The tools are heavily dependent on probability and statistics. The first tool, in fact, we call “probabilistic risk assessment.”
In simple terms, the Kodiak PRA decomposes scenarios into three primary factors:
- Scenario Exposure: How often does our vehicle encounter this type of operating scenario?
- Collision Likelihood: Given that our vehicle encounters this operating scenario, how likely is it that a collision occurs?
- Severity of Collisions: How severe would the collision in this scenario be?
The second tool is called “Breakpoint.” This tool feeds values into the three factors of the PRA:
BreakPoint deliberately injects realistic, time-varying errors onto the signals that flow through the autonomy system in order to ensure that the autonomy system still drives safely in both normal and extreme conditions. Better yet, BreakPoint actively guides its search adversarially, actively trying to drive our system into a “collision” situation, thus helping us discover autonomy failure modes. Then, BreakPoint tooling helps us estimate the risk associated with this failure mode, and this information flows directly into our PRA.
If you’re interested in solving the hardest technical challenge in autonomous vehicle development, you should read the blog post.
David Silver 