Ludicrous Mode, Range Mode and Chill Mode: Tesla’s various driving modes make for enjoyable driving suited to whatever situation a driver is in, be it the need for speed, more driving range or just cruising about.
Along with driver assist features known collectively as Autopilot – that helps make driving less stressful, and the ability to improve and add to these over time via “over-the-air” (OTA) updates – these are just some of the reasons that the Californian EV maker has built up such a solid fan base.
Autopilot can make the daily commute an experience to look forward to rather than dread, by taking the load off the mentally draining task of driving as numerous sensors and a swathe of algorithms recognise and react appropriately to common road hazards and obstacles, not least of all wandering pedestrians.
But there’s one feature, or shall we say character, of the Australian landscape that Tesla’s various driving modes and Autopilot do not seem to yet recognise: the kangaroo.
Driving around Australia is wrought with numerous challenges, not least because of our vast distances and challenging climate.
Add to that a rapidly bouncing native animal with a mind to abruptly change directions and absolutely no regard for road safety and you have an accident waiting to happen.
Take for example this event recorded by Tesla owner @Tesla in the Gong, where a lucky ‘roo barely misses being bowled over by his Model X.
— Tesla in the Gong (Yes, I am a Tesla 🤖 ) (@TeslaGong) October 12, 2019
Such incidents might be considered edge cases but examples like this caught via vehicles’ cameras by Australian Tesla owners in recent times has drivers such as TechAU’s Jason Cartwright asking Tesla CEO and co-founder Elon Musk to consider a new mode: “Roo Mode”.
Had an interesting encounter on the way home from basketball tonight. Skippy decided to jump across the road. I ended up braking myself, before the car did, actually not sure it would have avoided a collision.. definitely need 'Roo mode' @elonmusk #Model3AU #Tesla pic.twitter.com/lIVdkoqFN1
— techAU (@techAU) November 11, 2019
Is it feasible? Kangaroos have to be admittedly one of the more random and faster objects a car with driver assist sensors is likely to come across – the Australian Museum states that an eastern grey kangaroo (one of the more common ‘roos) was recorded once travelling at 64km/hr.
Sure, some cyclists may travel at such speeds, but kangaroos do not travel in a relatively predictable straight line (and nor do they indicate!).
Edge cases – so called “long tail” events – such as these don’t often happen compared to more common events such as a cyclist on city roads, but they still require attention if Autopilot is to make driving as safe as it possibly can.
We imagine that bouncing kangaroos is not foremost on the minds of Tesla’s Autopilot engineers but it does point out how different environments can throw up situations that engineers may otherwise neglect to consider.
As pointed out by Musk prior to the release of Tesla’s V9 software in 2018, it’s the “tricky edge cases” that make it important to get the positioning of Tesla vehicle sensors and repeaters just right.
Tesla Autopilot team is awesome! Got to make sure we iron out the details though. Long tail of tricky edge cases. Important to get front wide angle, B pillar & side repeater camera neural nets just right.
— Elon Musk (@elonmusk) October 2, 2018
But its not just the right hardware and algorithms that make recognising edge cases possible, it’s also Tesla’s access to large degrees of data thanks to its growing EV fleet.
As pointed out in October by MIT autonomous vehicle researcher and teacher Lex Fridman, over 625,000 Tesla vehicles are now actively delivering Autopilot data using Tesla’s Hardware version 2 and above (this includes Hardware 3 which is necessary for Full Self Driving, where as Autopilot will work on all vehicles with V2+).
— Lex Fridman (@lexfridman) October 12, 2019
According to Fridman who has projected these figures into miles driven by Tesla vehicles, it is thought that by 2020 there will be nearly 1.5 billion miles (2.4 billion kilometres) driven on Autopilot.
That makes for a lot of potential edge cases – but with a fledgling Tesla fleet in Australia numbering in the low thousands, the question for Australian country drivers is how many edge cases does Tesla need to be able to compute ‘roo recognition?
Correction: A previous version of this article stated that Autopilot miles were in the millions, not the billions.