At this point in any competition's lifetime, there are basically guys who have raced it many times, copied the published designs which worked, & mastered it. You're never going to win against those guys, but might be competitive. The general theme is you have to travel to Colorado & compete many times to gather experience in order to be competitive. They don't allow any significant practice on the course, before the race & they rearrange the barrels between each heat. Such a contest is no more practical than job interview coding challenges, especially if you're more interested in just being competitive rather than winning.
The only redeeming factor was how simple a winner's setup was, just a single RPM sensor for odometry & $5 Adafruit IMU for heading. There's no significant electronic or mechanical design.
His forward cam video is intriguing. It's very friendly to dead reckoning. The steering is real imprecise & the course is real small. The lion kingdom has been running semi autonomous robots for years at 10mph, for thousands of miles, & greatly enhanced the steering with frequency domane feedback.
The lion kingdom has no vehicle which could get over the climbing wall. The lunchbox is too big to fit between those barrels, but there's still the smaller, retired city robot. If the vehicle was as fast as possible for the steering to keep it on course, neglected wheel skid, it could probably still use dead reckoning with an alternative form of odometry.
Perhaps an extremely high framerate, downward facing camera with tons of artificial light would create effective optical flow odometry. It could track sideways movement, even when the vehicle was skidding sideways & track movement when it jumped off a ramp.
For the barrels, it just seems easiest to have it slow down, then detect when it hits something, then back up & try to go around it. Odometry needs to be real good to track position through that.
There is but 1 reason for pondering an extremely fast, robust vehicle for such a contest & that's military applications.