Nov. update: If Tesla's matured Full Self Driving would become available for licencing to other OEM's and start-ups should Aptera consider this?

Hello elfraty,

on 23 March Aptera announced that they will be using openpilot.

Here is the text of the announcement:

Quote

Carlsbad, CA, March 24, 2023 – Aptera is pleased to announce that it will integrate openpilot™, the number one ranked driver assistance system by Consumer Reports, in its solar electric vehicles. This move will allow Aptera drivers to enjoy road trips with openpilot™’s advanced features including adaptive cruise control, lane centering, and forward collision warning. The driver monitoring system will ensure that drivers remain attentive while using the system.:So it is clear where and how the journey is going for the time being

So it is clear where and how the journey is going for the time being

We drive a Tesla with "FSD Autopilot" ourselves, and I appreciate the assistant, but I know its limitations.

The IONIC 5 has the HDA 2.0 (Highway Drive Assistant) installed. It only works on the motorway and can absolutely hold a candle to Tesla. Tesla is clearly inferior in some things, but slightly better in others.

I would like Aptera to continue working with openpilot.

--------------------------------------- Deutsch --------------------------------------------------------

Hallo elfraty,

am 23 März hat Aptera bekannt gegeben, dass sie openpilot nutzen werden.

Hier der Text der Verlautbarung:

Quote

Aptera freut sich bekannt zu geben, dass es openpilot™, das von Consumer Reports an erster Stelle bewertete Fahrerassistenzsystem, in seine Solarelektrofahrzeuge integrieren wird. Mit diesem Schritt können die Fahrer von Aptera ihre Fahrten mit den fortschrittlichen Funktionen von openpilot™ genießen, darunter ein adaptiver Tempomat, ein Spurhalteassistent und ein Kollisionswarner. Das Fahrerüberwachungssystem stellt sicher, dass die Fahrer bei der Nutzung des Systems aufmerksam bleiben.

Damit ist klar wohin und wie vorerst die Reise geht.

Wir fahren selber einen Tesla mit "FSD Autopilot", und ich schätze den Assistenten, kenne aber ich die Grenzen.

Der IONIC 5 hat den HDA 2.0 (Highway Drive Assistant) eingebaut. Der funktioniert nur auf der Autobahn und kann Tesla absolut das Wasser reichen. In einigen Dingen ist Tesla deutlich unterlegen, in anderen etwas besser.

Ich würde es gut finden, wenn Aptera weiterhin mit openpilot arbeiten würde.

Same here. Using Teslas since 2016 and collecting more than 200.000 electric kilometers, I loved its autopilot until around 2017/18 or so. From then on it got more shaky, needed more and more supervision and created uncomfortable situations more than in the early versions. So I completely agree that it has significant limitations.

Nevertheless I catch myself enjoying myself driving the car more again. So it's ok for me that the first apteras would not have openpilot at all or only limited functionality (e.g. due to "only" two cameras).

Quote from e-economy

200.000 electric kilometers,

WHOW !!!!!

Quote from elfraty

should Aptera consider this?

in one the interviews or webinars Chris mentioned the power draw consideration of autopilot computers that they have to take into account ... so I guess one important metric that no one is yet talking about is going to be "performance per watt". I just don't know yet how to define "performance" in terms of autopilot

Quote from j.pecar

I just don't know yet how to define "performance" in terms of autopilot

Its a good point, all actual systems in cars are more-less a gimmick than an autopilot!

If autonomous driving capability cat 4 & 5 would be realized I would say, current draw would not be an issue, because to the benefit! Further I doubt a camera system (how many redundant cameras ever) without active sensors will achieve such an evolution. Also Tesla not! Open Source system (OpenPilot) would be good for for system evolution, but not for getting nominated persons in responsibility after accidents in CAT 5 mode.

Besides law regulations (only CAT 4 prepared and allowed in Germany):

(deutsch) https://bmdv.bund.de/SharedDoc…zum-autonomen-fahren.html

and safety rules and conventions for AI decisions (still have to be defined) and aspects it will need data recorders and cameras to unveil accidents. It cant be only a simple dash cam with autopilot add-on like OpenPilot. Its much more and will definitely require massively processor capability acting in a 5G/6G network system negotiating with peripheral supporting devices. Devices of public infrastructure as well the devises of all traffic around.

I afraid its quite difficult and expensive for cat 4 & 5 to add and wired all that required systems with sensors, band transceivers afterwards in an Aptera. I also dream much, but have been realistic too. Knowledge stand today 2023 in Germany, and for the US market easier to establish. For the EU only after further investments and much updates as like all others manufactures have to do.

November update:

For the enthusiasts of self-driving capabilities (I as well), here a video from the Free Power channel with title "Aptera Will Have FSD !", but he sets Tesla FSD system equal with OpenPilot just because AI driven technology. Its not as we know, but we are not going into details now.

Despite OpenPilot is not considered as a legal option coming to Germany, but might sneak in, because possibly will be allowed in countries around Germany. Here everybody needs to understand the legality in countries is not primarily the technology question or it's technical evolution only, it's the question how will AI fully hit countries societies in ethics. At least the question is to be answered, does AI decide by algorithm who will survive in an inevitable accident, where only the question left or right, and where people standing.

Or the society decides to give room to this technology to evolve, getting better and better and result in less total losses in lives and better statistics. In comparisons to the faults of human decisions annually cause higher kills or injured persons.

For instance, in direct question to example above, at left side a child, right side an old man, in first objection most people in Europe say in this scenario the AI should choose the old man.

When the AI and data transfer systems will be capable at a future day to distinguish between within milliseconds.

Here most French people spontaneously, and some others, said no, the AI should use a unequivocal random algorithm to decide fully by chance and the technology should not play god or make any technical selections.

On top in Germany, all have to be recorded, fully evaluated and according German law suit actually a physically person have to be able responsible for it. The OEM of software or the integrator of this software. What means, CEO of Comma ai or Chris/Steve of Aptera, when allowed also in Germany or the driver personally at usage in Germany. For the last unfortunate in this example, German prosecution for sure will investigate, judge and punish. That will require technics in cars in Germany to record, like an aircraft black box, to assure nothing can be deleted.

You all remember the occasions in US about suddenly accelerating (automatic transmission) cars? Causing millions of USD compensation in law suits. It will be interesting how US law in will cope with that in future with FSD, quite similar and comparable.

And now following fact drives the arguments above irritating by announcements of the European Union for new requirements to Intelligent Speed Assist (ISA) systems as an update of the “General Safety Regulation” (GSR).

For example here the EU wants or better tries to regulate technical requirements in new cars with FSD starting mid of 2024.

SLIF = Speed Limit Information Function (Verkehrzeichenerkennung)

SLWF = Speed Limit Warning Function (Geschwindidkeitsbegrenzungs-Warnfunktion)

SFC = Speed Control System as alternative for SLWF. This harsh function decelerates automatically the car if exceeded.

Following german video (sorry about having in German only):

Das TEMPOLIMIT im Auto kommt! 2024 wird diese "ISA" EU-Richtlinie PFLICHT für ALLE Neuwagen!

And describes at min 9:50, this functionality has to be assured full reliable working also via crossing country borders and by a click on a control panel or fully automatic. This is not a regulation only

associated with FSD, it’s also a new regulation for the normal speed limiter functions in travel assist systems.

My opinion here:

These requirements written down 2019, and without clues about what AI is, by announcements will be set active 07/2024 for all new homologized cars on EU-Roads with speed limiters.

Which means game over for a long time for already designed FSDs and driving assist systems on EU-roads because these cannot be realistic implemented. The German vid shows where are the traps, but also makes not all necessary conclusions. I have to find more information how this is really associated with FSD, besides any FSD have to use this function. A functional explanation, not really must comply with authority intent. The originators of this regulation had no clue about what is feasible with modern technics, and raise bars too high for the actual technology be capable soon to achieve the required mistakes resilience within functionality.

A question still pending to clarify is, why a FSD system has to be classified during homologation together with vehicle type? I case of OpernPilot we know, its independent of car type optional implementable and needs only a connection to CAN-Bus information.

GSR/ISA requirements unfortunately crosses out, speed limiters can vice versa rely or feed on FSD supplemental given information. Here the area-information, traffic sign recognition!

Looks like a disaster, but GSR ISA requirement have to be interpretated differently, thus FSDs can legally implemented.

What a mass, at the time being, I see only Tesla and Openpilot with their AI driven FSDs in future be capable to master those requirements by AI functions, because bad visibility, dirt, glare can prevent photo evaluations and need to be resolved by AI pattern comparisons as a backup. Completely without AI this is actually nearly impossible to achieve with conventional technologies in normal cars, or only with much higher costs for new sensors and much more capable evaluation software accompanied with higher onboard hardware requirements needed!

By chance and may only possible only together, because both approaches have massively disadvantages in deep learning or preparing and providing updated map data, also to cope with temporary changes. E.g., at road construction work! This may only can achieved in connected systems to use both approaches with overlapping backup functionality but then also with new expensive problems to resolve.

At the moment I pessimistic about FSD because, this topic is only the speed limiting function any FDS need to use. Here speed limit traffic signs, there a Stop-Sign recognition. All isolated problems have to resolved, and there are many others. If this all will be regulated in advance, no evolution in software, testing on real daily scenarios will be possible. That is crucial for such systems to get big data for pattern improvements. That is crucial for such systems to get big data for pattern improvements.

I see a big gap between full accomplishment the authorities require, between system evolution provides at this time. The transition period seems to be overregulated thus the gap cannot be closed.

Self-driving got a show stop from EU with that regulation and may a reason why many manufacturers to surrender.

Best regards

André