Robo-Cars Need Relationship Map 2017-03-29
Even if it doesn’t quite rival the messy relationships in a daytime soap opera, autonomous vehicle partnerships are definitely part of a tangled web.
The Daimler – Bosch deal announced earlier this week, however, has brought some clarity. Daimler and Bosch said Tuesday ([**]pril 4) that they’re partnering to accelerate the production of "robo-taxis.”
The real significance of the news, however, is that all three -- Daimler, Mercedes Benz (owned by Daimler) and Tier One Robert Bosch -- have chosen Nvidia as their autonomous driving platform partner.
Mobileye vs. Nvidia
In fact, when it comes to the platform battle over Level 4 and Level 5 cars, the automotive industry today is split in two: you are either with Mobileye (soon to be part of Intel), or with Nvidia.
[**]nd Nvidia has clearly shown it is gaining momentum.
[**]sked about the Damiler-Bosch announcement, Phil Magney, founder & principal advisor at Vision Systems Intelligence, told us that Daimler must be “encouraged with Nvidia’s roadmap for both [**]I and solutions for highly automated driving.” Notably, Mercedes announced a plan to develop [**]I-based self-driving capabilities with Nvidia even before the Daimler-Bosch partnerhips happened, he explained.
![Relationship Map among chip vendors, tech companies, Tier > <div style=]()
Relationship Map among chip vendors, tech companies, Tier Ones and car OEMs (Source: EE Times)
On the other hand, an opposing camp built around the BMW/Intel/Mobileye alliance was announced last summer. Last month Intel announced the acquisition of Mobileye. Separately, Intel and Mobileye last year picked up Tier One Delphi as their partner.
Especially in recent months, among high-profile autonomous car development announcements with prestigious German car manufacturers, Nvidia is stealing Mobileye’s thunder.
But look closely, said Egil Juliussen, director research, infotainment & [**]D[**]S at IHS [**]utomotive. Mobileye might be “way ahead of Nvidia,” when considering the number of OEMs and Tier Ones already deeply aligned with Mobileye.
What Mobileye's annual report tells us
Juliussen is referring to Mobileye’s unprecedented dominance in the [**]D[**]S market. The company has an almost unfair advantage when it comes to the installed base of the company’s computer vision modules. [**]ccording to Mobileye’s Form 20-F SEC filing, Mobileye's modules are “installed in approximately 15.7 million vehicles worldwide through December 31, 2016.”
Mobileye said, “Our technology is available with 21 OEMs. Furthermore, our products have been selected for implementation with more than 25 OEMs.”
But of course, this is all [**]D[**]S. [**] big question, said Juliussen, is “whether Mobileye can transfer their relationships to the autonomous vehicle platform.”
Mobileye’s SEC filing, however, indicates that they are not doing badly in coming design wins. The company has design-wins from five OEMs for Level 3 autonomous driving in the pipeline. It also has wins from five OEMs for Level 4 autonomous driving.
Many auto industry observers believe Mobileye can effectively convert from [**]D[**]S to autonomous vehicle design.
Part of the reason is that the Israeli team hasn’t stopped innovating. The Mobileye team, now a division of Intel, led by [**]mnon Shashua, a computer science professor at the Hebrew University in Jerusalem, co-founder and CTO at Mobileye, has been already put lots of resources into developing substantive technology components that go way beyond [**]D[**]S.
![[**]mnon Shashua, co-founder and CTO of Mobileye](res/201704/08/auto_96.png)
[**]mnon Shashua, co-founder and CTO of Mobileye
When Intel announced the acquisition of Mobileye, Magney observed, “I think Intel believes Mobileye’s team is further along in the development, not only vision but other technologies like behavior (Driving Policy) and localization (REM).”
[**]ccording to Mobileye, REM is “is an end-to-end mapping and localization engine for full autonomy.”
The solution consists of three layers: harvesting agents (any camera-equipped vehicle), map aggregating server (cloud), and map-consuming agents (autonomous vehicle).
Considering the plethora of camera modules already installed in the “harvesting agents,” Mobileye has a lot to gain by promoting REM. [**]nd the company is making sure that those OEMs who are already hooked on Mobileye’s computer vision modules feel the same way. Data collected by these “harvesting agents” is paving the way for the eventual BMW/Intel/Mobileye autonomous vehicle platform to succeed on the road.
Multiple development programs
It’s hard to predict, however, whether these publicly announced partnerships will stay intact in the long run. Some car OEMs might simply want something quick so they can show a cool autonomous car demo on the road by 2020 — as promised — with little intention to make money from it.
While others might be gunning for autonomous vehicles, they feel pressed to explore a new “robo-taxi” business model.
Magney explained, “Urban mobility platforms probably represent the best ROI in the near term because the business models for shared mobility are so well established.” He said, “Furthermore, these vehicles allow their makers to deploy within a ‘somewhat’ constrained environment compared to full L5 automation anywhere.”
It’s also important to note that vehicles deployed for these applications will have “a heavy [**]V ([**]utomated Vehicle) Stack. Very costly, multiple redundancies, fail safe, lots of data collection and OT[**] (Over the [**]ir) capabilities,” as Magney noted.
This is good for gaining experience but the danger is an autonomous vehicle that ends up as a science project with little practical value.
Meanwhile, car OEMs are known to run multiple development programs in parallel. Consider Mercedes. Before its parent Daimler decided to partner with Bosch, Mercedes-Benz reportedly had two engineering teams working on autonomous vehicles. Reuters reported, “One took an evolutionary approach, upgrading the capabilities of conventional vehicles, while the other team took a more radical approach to the car's design.”
Magney believes it likely, given especially at this stage, “they have to examine different approaches.”
Who’s in bed with whom
[**]s Juliussen said, there’s the BMW/Intel/Mobileye group. [**]s Mobileye’s SEC filing indicates, this group has signed up at least five OEMs for Level 4 cars, he explained.
Then, there’s the Nvidia’s platform. Besides [**]udi and Daimler, Nvidia has partnered with both Bosch and ZF – two Tier Ones. “They are significant,” said Juliussen, because they can get more automotive OEMs.
In contrast, both GM and Ford, the two large car OEMs in the United States, are pursuing a similar strategy by acquiring automotive software companies. GM acquired Cruise [**]utomation a year ago, while Ford bought [**]rgo [**]I two months ago, each OEM shelling out $1 billion for the acquisition. Neither has announced its hardware choices.
Waymo announced deals with Honda and Fiat Chrysler, under which each OEM supplies vehicles to Waymo. It is not known if the Honda-Waymo partnership will evolve any further.
Toyota launched the Toyota Research Institute (TRI), which is investing over a billion dollars in robotics and artificial intelligence over the next five years, but has not announced either a hardware or software platform.
Gil Pratt, a former MIT professor who heads the year-old TRI, most famously said during the CES 2017, even with huge recent advances in artificial intelligence ([**]I), "We are not even close" to Level 5 autonomous cars. While some people felt his view was extremely conservative, others found Pratt brutally honest and realistic.
Magney said, "The Toyota approach is very pragmatic and probably aligns with the realities of an automaker that plans to sell millions of cars to consumers before we have mass shifts in ownership."
While details of their own highly automated driving platforms remain unknown, Nissan, Hyundai and Kia are all using Mobileye for their [**]D[**]S cars, according to Juilussen.
It should be also noted that Uber and Daimler earlier this year struck partnership for self-driving vehicles. Magney said, "You have to acknowledge the potential of Uber -- if they get the technology right they can roll out through their existing mobility platform where they have huge advantage."
Last but least, there is Tesla. [**]fter its fallout with Mobileye, Tesla is currently working with Nvidia. Magney said, "Tesla deserves a lot of credit for what they have accomplished -- they developed the right architecture from the beginning and started to enable automation features with software."
Options for startups and OEMs
IHS [**]utomotive’s Juliussen explained several different ways those interested in autonomous cars can enter the market.
If you are a startup, said Juliussen, “you source EV batteries, focus on luxury vehicle models and buy an autonomous software platform from someone else or design your own. Then, get a contract manufacturer and have them build a car for you.” This is a model followed by companies like Zoox, Waymo, or potentially even [**]pple, he noted.
If you are an OEM, there are three options, Juliussen said. “First, you develop your own software platform -- either through the acquisition of an automotive software company or via your internal team.” This is the approach taken by GM and Ford.
Second, “you could follow a more normal strategy by hiring a Tier One” to design an autonomous car, he said. You can, for example, hire Delphi to help you build a Mobileye-based autonomous vehicle platform, he noted.
For many OEMs, this is a tried and tested model. [**]sked what Daimler was seeing in Bosch that Daimler couldn’t do it alone, Magney said, first and foremost, “Integration.” He explained that Bosch brings deep integration skills and know-how to make a system truly production-ready. Magney added, “Bosch has lots of ECU experience as well as supporting domains with the [**]V Stack such as sensing.” But in the particular case of Daimler’s partnership with Bosch, Magney suspects the decision was driven by “Daimler’s desire to go with Nvidia’s architecture and supporting frameworks.”
[**]s for smaller car OEMs with not enough resources, Juliussen suggested a third option. “They can get a software platform from the third party and get a contract manufacturer to build it.”
Or, they can forget about autonomous cars altogether and focus on the development of [**]D[**]S vehicles.
Regarding the brain chip for autonomous cars, competition is currently shaping up as a two-way battle. But this is a market in its infancy. Just this week a new player, Mentor, revealed a centralized autonomous vehicle platform with architecture squarely focused on raw-data sensor fusion, opening up another option Tier Ones and car OEMs can consider.
Magney summed up: "The building blocks for automation are out there as the push from the tech community has shown. The real challenge is the integration into a total automotive ready platform. There are so many domains within the [**]V Stack (perception, behavior, control and safety) and nobody owns all those pieces. This is why we see a flurry of activities related to centralized domain control for highly automated driving."
