Last Friday the guys from Bosch came to campus invited by the UC Berkeley EECS School.
Bosch, like BMW or Mercedes, has a Research Centre in Palo alto where the engineers are creating the “Vision and the Roadmap” for automatic driving. The Centre participated in the Urban Challenge organized by DARPA and since 2010 is prototyping systems for the project.
When speaking about “Automated Driving” one have to distinguish between “Supervised by the driver” (some technologies are available today such as Park Assist, Integrated Cruise Assist and others, like Highway Assist, are progressing fast), “Highly Automated” (Highway pilot) with reduced driver supervision and “Fully Automated” (Auto pilot).
When we will see fully automated cars on the market? According to Bosch, it is likely that in 2020 we will see the first commercial prototypes of “Highly Automated” cars. No date for “Fully Automated” cars can be forecasted today…
But what is missing ?
- Surround sensing… in all circumstances !!!
- Safety and security
- Very precise and dynamic map data
- Highly Fault Tolerant System Architecture (what happens if the battery dies???)
Let’s examine briefly some of those aspects…
Today, 360° surround sensing is possible with the use of radars, sensors and cameras but there are issues in special circumstances … what happens tunnels, low sun or with some materials like timber f they are transported by trucks?
What is missing, among other things, is what is called “Third Sensor Principle” in excess of radar and cameras. Sensors that work in real time, asynchronouly, using probabilitic algorithms and computationally very efficiently with supervising systems that are able to decide in cases of conflicting information. In fact, a new generation of sensors…
Dynamic Map Data
Today, most of the map data we have in our GPS is mostly static. What is needed is absolute localization data on maps with dynamic layers, much more precision and SLAM (Simultaneous Localization and Mapping).
The driver has to be monitored to detect distraction, drowsiness, Health state, etc. Identification and adaptive assistance are also necessary and also the ability to return control if necessary and this is a Key Element for this part.
Protection against technical failures by means of redundancy in the steering and braking systems (some elements like assisted steering, ESP HEV and iBooster exist already today, particularly in electric vehicles).
One important aspect of security is quality control and testing for release. In traditional cars the quality control is done statistically but this method will not be feasible for the testing of fully automated driving vehicles. It is estimated that the number of test hours would be multiplied by a factor of one million. New release strategies are needed with a combination of statistical validation and new qualitative design and release strategies for individual components and the full integrated system…
Currently laws regarding traffic and car driving are enacted at National level. However , there are two international conventions, Geneva (1949 UN) ad Vienna (1968), on road traffic… the problem is that some countries have ratified one and not the other… or neither of two !!!!
Needless to say that , like in many fields of technology, legislation does not reflect easily and quickly the technical progress….
To illustrate it, it looks that in the Vienna ( or Geneva, I do not remember..) Convention it is stated that
“Every driver shall , at all times, be able to CONTROL his vehicle or GUIDE HIS ANIMALS”…. one can imagine how modern thsi rules are when it speaks about.. ANIMALS…
The key here is the meaning given to the word “CONTROL”… if taken literally there is no possibility of “Automated Driving”… but what if CONTROL would mean “SUPERVISION” ?
It looks that the state of California has accepted the “testing” of such kind of vehicles based on well justified request and with “certified” people… at least research can continue… well done.. I am sure that Google has something to do with this… ;-)))
One of the questions that is often raised is “What will be the User Experience (UX)”?
What the driver will feel?, Emotions? Transition of control back to the driver? ….
It has become clear that the automotive industry has become a hardware/software industry , the mechanics is still important but the car is full of IT systems that have to work in an integrated way and with very fast response times.
This is even more important in the case of full automated driving and the software has to be fault tolerant, secure and very efficient.. …imagine the power of the embedded processors to take reactive action in milliseconds to determine the trajectory… and at the same time, but a little bit slower, say in seconds, decisions about the manoeuver…
Bosch displayed a video that illustrated their vision for Highway Pilot by 2020. I will point to it as soon as that upload to their web site because is very interesting…
The meeting, as usual, ended with a request to EECS students to send applications for jobs/internships at Bosch Research Centre in Palo Alto..
At the end of the meeting, I asked some questions
Question: “There are initiatives by Google in this field and recently the press has published that Apple might have 1000 engineers working on the subject… are you working with them?
Answer: “We are not authorised to speak about the collaboration with partners”…..
Question; “Bosch is not a car manufacturer, what is your business model for this technology?”
Answer: “Bosch is a manufacturer of components or complete systems for traditional or electric cars. We are going to continue with the same approach”
Question: “If there are several suppliers of components or subsystems on the market, and the car manufacturer decides to have multiple suppliers, they will have to work together in a mission critical framework. Standards will be needed; what is the current situation?”
Answer: Today there are standards at low level that allow communication among components and subsystems. Higher level standards will probably be needed in the future but which and when is difficult to say today. having said it, we believe that the automotive industry will find the necessary agreements to ensure interoperability at least at some level”
One of the students asked a very interesting question: “Will automatic driving avoid today’s typical accidents? There will be more? Or Less?
Answer: “100% safety will never exist. The first accidents, when the full automated driving will be available, will produce big headings on the press. This is not new: the introduction of safety belts, airbags ,etc in traditional cars was criticized at the beginning; with the time and the technological progress the criticism has mostly disappeared; nobody would accept today cars without those safety elements and legislation is enforcing them . As far as the number and type of accidents are concerned, Bosch believes that there will be less accidents and those will probably be somehow different to the accidents in the case of human driving…”.
Very interesting conference on a new subject…… at least for me…
I spoke to the Director of Research at the end of the conference and we agreed that I will visit the Center when I will go to Palo Alto, Mountain View or Menlo Park for other meetings in March or April
Stay tuned for more…