Autonomous FSAE?

[Claude Rouelle]

Some people won't like what follows but hey you won't make omelets without braking eggs...

SAE created the FSAE many years ago. Then many other competitions were also organized with more or less success - most often more. Formula Student Germany created Formula Student Electric and other organization followed. SAE became a follower.

Now Germany and, I think, China do organize an autonomous Formula Student competition, which does make perfect sense as it where the automotive industry is going.

How come this competition "Removing bumps on the path to fully automated driving"
http://www.nxtbook.com/nxtbooks/sae/17UPD07/index.php?partner=35314341433341303735423544353632 38443635363232334343413035354131&startid=4&lre=1%3A%3CENCRYPT_CUST_NUM%3E&mkt_tok=eyJpIjoiTW1NeE9EbGhZelZtTTJVMCIsInQiOiJFcV lcLzVSVGxYQ3phU21pZXZ1dFpOZFE5UEFrMkQ2SWlDWlpaN3Ra aXNjZDRTZVpFSlpnMHEydzNPTFJDY3g2NVdCVEttRXhpXC9BeU 5xbHFJUWVyRFdJUWFQXC95TlBnXC9YTndiS2dPYnZUSzVCXC9w WlQxNjZjem03MnBqWU9tV2ZZIn0%3D#/4
is not integrate in a series of international competition that counts close to 1000 universities in the world?

In the interest of the students as well as and the US and global automotive industry what does it take for SAE for make FSAE great again?

[tim_pattinson]

I have to agree here. Irregardless of how AutoDrive, the SAE comp, is run. it's an invite-only event.
The end result is there are only eight teams registered, compared to 15 at FSG with two on the waiting list (5 withdrawn)
From this page: http://students.sae.org/cds/autodrive/event/


This newly established, three-year autonomous vehicle competition will task students to develop and demonstrate a full autonomous driving passenger vehicle. The technical goal of the competition is to navigate an urban driving course in an automated driving mode as described by SAE Standard (J3016) level 4 definition by year three.


The competition will consist of up to 10 teams from select universities that were invited to participate in the inaugural competition. These foundational teams will be announced April 2017. The first of three yearly competitions will take place the following spring 2018.

Is it too much to expect teams to come up with an autonomous concept in a year or so? We'll see next August. But the idea of a 3-year cycle is questionable.
I'd say 95% of team members last less than three years. Plus how happy will universities be if the team has nothing to show after a year? Two? It's hardly ideal for students either - You join the Formula team and hopefully see your design work in about a year's time.
With this, it's uncertain, and the feedback loop from crap design to crap performance is a lot longer, meaning students will learn less and the competition will always be behind FSG.

edit:
Also, what's wrong with racecars, and where are teams supposed to test a road car, in a realistic situation, where it won't launch into the nearest fence or wall?

[Bemo]

In my opinion a three year cycle cannot make sense for any Student competition. It's the essence of FSAE that every Team member experiences the entire cycle from the early concept Phase until the testing and Evaluation with the actual car. A three year cycle will mean that most of the students of the beginning won't be able to participate at the actual competition. How should that motivate anyone?

I can't understand why you would organise such a competition as an invite-only Event. If there is a motivated Team at any University they should come and present their efforts...

[Claude Rouelle]

I do agree with Bemo; the competition should be as open, as international and as less exclusive as possible

[DougMilliken]

Claude,
I looked briefly at the SAE/GM event and it doesn't seem anything like FSAE (except that cars are involved). The 8 teams are given Chevy Bolt electric cars and their task is to turn them into self driving cars. Since GM is giving away cars, it makes sense to me to have some initial vetting--so their investment isn't wasted on a team of dreamers that don't have the resources to make useful progress.
Again, I didn't study the rules in detail, but this looks like a software contest, with integration of purchased sensors. Nothing like the sequence of engineering analysis, mechanical design, build, test & correlate in FSAE. Do you have other reasons to compare this to FSAE/FStudent?
-- Doug

[noah]

I hope and assume SAE does not see the GM autodrive as a replacement for Autonomous FSAE in the US. My hope is that FSAE will create an Autonomous competition in the US within the next 2 years so we don't have a repeat of the massive catchup that American electric teams are still clawing back.

-Noah

[Claude Rouelle]

My point is not to criticize GM action that has merit. My point is that such action could have been initiated long time ago on larger scale as part or FSAE / Formula Student competitions in the US.

Just a question of choice leadership because I can tell you that in term of money, the money is there.

Maybe it is time for a Formula Student USA of the students, by the students, for the students. With the support of universities and companies.

Claude

[DougMilliken]

My hope is that FSAE will create an Autonomous competition in the US within the next 2 years so we don't have a repeat of the massive catchup that American electric teams are still clawing back. -Noah
Maybe I'm dumb, but I don't see the connection between racing and self-driving cars, a few thoughts:

+ At the local/amateur level, racing is about mastering a machine. This can be addictive and certainly very satisfying for the driver and friends that make up the team. FSAE adds in the challenge of creating the machine, with the reward of engineering/project-management experience and the fun of driving the car.

+ At the pro level, racing is a sport/spectacle for fans and a big TV audience, with sponsorship primarily by corporate marketing (some car companies also use racing as training for new engineers). For engineers in racing, it's a high pressure job with huge time commitment, which can be rewarding (or can cause early burnout).

+ After watching some of the early trials for the first DARPA Grand Challenge, I have no further interest in spectating at a race between self-driving cars (or watching this on TV).

+ If you are interested in self-driving cars as an engineering challenge, there are already many types of "battle bots", and perhaps these will become more autonomous as time goes on? (I don't follow these events closely.)

+ Cutting through all the hype, self-driving cars are some kind of cross between a cheaper taxi/Uber and faster public transportation. The more expensive projects that I know about focus on a smooth ride, so the passengers can work/read without getting motion sickness. Boring as watching paint dry (as long as they don't fail!)

Where do racing and self-driving cars intersect?

[noah]

Maybe I'm dumb, but I don't see the connection between racing and self-driving cars, a few thoughts:

+ At the local/amateur level, racing is about mastering a machine. This can be addictive and certainly very satisfying for the driver and friends that make up the team. FSAE adds in the challenge of creating the machine, with the reward of engineering/project-management experience and the fun of driving the car.

+ At the pro level, racing is a sport/spectacle for fans and a big TV audience, with sponsorship primarily by corporate marketing (some car companies also use racing as training for new engineers). For engineers in racing, it's a high pressure job with huge time commitment, which can be rewarding (or can cause early burnout).

+ After watching some of the early trials for the first DARPA Grand Challenge, I have no further interest in spectating at a race between self-driving cars (or watching this on TV).

+ If you are interested in self-driving cars as an engineering challenge, there are already many types of "battle bots", and perhaps these will become more autonomous as time goes on? (I don't follow these events closely.)

+ Cutting through all the hype, self-driving cars are some kind of cross between a cheaper taxi/Uber and faster public transportation. The more expensive projects that I know about focus on a smooth ride, so the passengers can work/read without getting motion sickness. Boring as watching paint dry (as long as they don't fail!)

Where do racing and self-driving cars intersect?
Doug,

The longer I participate in FSAE, the more it becomes apparent that less and less of the students involved are participating for the motorsports aspect. I am in the minority, but the majority of the team has no intention to work in motorsports or possibly automotive in general. The vast majority of the students these days have joined FSAE as an engineering competition, not a motorsports or automotive competition. They know that many employers value FSAE over everything else when it comes to hiring choices.
Say my team has roughly 60 active members. 3, maybe 5 tops want to work in racing, myself included. 20-30 want to work anywhere in automotive sector. The remaining 25 or so don't care whether they work in automotive or aerospace or kitchen appliances. For the 55 members of my team who are in it for a job, one of the largest growing sectors for them to join is the self driving sector, and a FSAE competition will give them the fast paced environment they need to be prepared for the industry. Plus the University will love it and it will make my job a hell of a lot easier to find companies willing to support our teams so I can get the funds I need to keep our driven cars racing.

[noah]

Also,

You need to remember that nobody is going to spectate an FSAE event for the pure racing action. As much as I love seeing the cars go round, I love to see it because of my involvement in it and background knowledge of my team and the teams we are facing and the challenge and engineering rational each car is based on. Watching driverless cars go around will be exactly the same. please take a minute to watch this video AMZ just put out: https://www.facebook.com/amzracing/videos/10155033171416107/
If seeing that car go around that fast doesn't make you at least a little excited, then I am sorry. A driverless car can also hypothetically go faster than a driver ever could, especially with some of these crazy 4wd torque vectored monsters from Europe. My guess is the ai can take advantage of the turning ability in ways a human never could.
https://www.facebook.com/amzracing/videos/10155033171416107/

[Z]

... what does it take for SAE for make FSAE great again?
Claude,

Surely the answer to your question comes from the same source. Namely...

DRAIN THE SWAMP!!! :D

Z

[Bemo]

Claude,
I looked briefly at the SAE/GM event and it doesn't seem anything like FSAE (except that cars are involved). The 8 teams are given Chevy Bolt electric cars and their task is to turn them into self driving cars. Since GM is giving away cars, it makes sense to me to have some initial vetting--so their investment isn't wasted on a team of dreamers that don't have the resources to make useful progress.
Again, I didn't study the rules in detail, but this looks like a software contest, with integration of purchased sensors. Nothing like the sequence of engineering analysis, mechanical design, build, test & correlate in FSAE. Do you have other reasons to compare this to FSAE/FStudent?
-- Doug

That makes it even more pointless to run it in a three year cylce, as there is no work to do for getting the car running at all...

In Formula Student driverless the Task is actually very similar. Teams also convert existing running cars into autonomous vehicles. But they don't Need three years (why should they...)

And the limited number of those cars is also no reason for an invite only program. You could run a concept competition which is open and the best Teams of that can implement their concepts in the actual cars.

Besides that I can only agree with Noah. FSAE is not a racing series. Therefore you should not argue about things like that as in regular racing series. The Major Point of FSAE is to give students a platform to gain practical experience in Engineering. Why should it be wrong to implement a field which is rapidly growing and promising great Job opportunities?

[JulianH]

I agree with the statements made here:
1) We need one-year cycle projects. Histroy showed that all teams that were running 2-year-cycles did not perform close to the level of the "regular" teams. 3-year student projects do not work. You can do that for PhD students or something like that.
2) I am not a big fan of Autonomous motorsports. I don't see the point in those "Robo Races" that Formula E started to do but that is just personal. Maybe car OEMs will like to showcase their skills in Autonomous racing (see the Autonomous RS7 from Audi 1-2 years ago). I don't know if the fans will approve.
3) I agree with Noah: FSAE is not about "getting to know motorsports". It is in the end an engineering and project management competition and I think adding an extra layer of complexity is super fun.

A couple more points:
- I was really quetioning the possibility of a -somewhat- successful Autonomous FSAE class; so far my successors in Zurich seem to be on a good path to create a somewhat fast car, so I assume it could be fun to watch in August
- I am a bit sceptical if we can "roll this out". So that teams build Autonomous cars from scratch, etc. Bemo, I think the long term plan should be more than "transforming old cars into Driverless". Right now the cars are too heavy for the 10lap trackdrive due to too many batteries, and probably also too powerful.
I don't know if we have enough teams and resources to have 3 independent competitions in FSG...
- I am also a bit afraid that the "gap" between top teams and "regular" teams get incredibly big due to this new challenge.
Right now we have 10-15 teams in Europe that really kick ass, that create awesome cars that are amazing to look at and watch and we probably have >100 teams that well are somewhat stuck in 2009, do not really develop...

So let's see how this will play out :)

[Babenaldt]

Being from a University that was "lucky" enough to be chosen for the Autodrive challenge, I've noticed one misconception about the competition. It is a yearly competition where you bring the same car for three years, not one competition every 3 years.

[DougMilliken]

Noah (and All) -- interesting to hear the makeup of your team. This is quite different from ~2000 when I started as a design judge. The students I met then were mostly interested in some aspect of the auto/racing industry. I have heard similar comments about recruiting team members where the FSAE teams were competing for people with a solar car team (or other student vehicle project). I still claim that these two driverless events are very different competitions and shouldn't be compared. While there may be some team member poaching, it doesn't seem like these would attract the same kinds of people? Below are some further thoughts on comparing the two events.

Reference to your second post, I tried to say the same thing in my first and third bullets. These are non-spectator events, interesting for participants, same as most amateur racing. This includes autocross which is the stated market for FSAE cars (we could have a separate discussion about autocross as the basic justification for FSAE, but I think some other threads already exist.)

FSD (driverless): I found this description of the event,
https://www.formulastudent.de/pr/news/details/article/autonomous-driving-at-formula-student-germany-2017/ and also the rules from, https://www.formulastudent.de/pr/news/details/article/rules-2017-v10/

From the video you linked the car was following a fully double-sided cone course. It appeared that they were staying on the centerline between the cones. During the event they don't have to pay attention to the rest of the environment, for example, rule,
> D2.1.2 [DV ONLY ] There will be no flag signs for DV in autonomous mode.

And the driving task is also simplified,
> D8.1 Trackdrive Tracklayout
removes the slaloms from the list of features for the regular endurance course.
> D4.1.8 Independent of the weather, the track conditions will be made artificially “wet” (not for DV).

For me, watching this video doesn't depend on the vehicle, it could be this re-purposed FSAE car, a little RC car, a battle bot (with onboard computing and sensing) or, with a scaled up course, a full sized car.

SAE AutoDrive Challenge: description here, http://students.sae.org/cds/autodrive/event/
> ... The technical goal of the competition is to navigate an urban driving course in an automated driving mode as described by SAE Standard (J3016) level 4 definition by year three.

Level 4 is here, https://www.sae.org/news/3544/
> Level 4 – High Automation: The driving mode-specific performance by an Automated Driving System of all aspects of the dynamic driving task, even if a human driver does not respond appropriately to a request to intervene

Dealing with highway situations and traffic is orders of complexity harder than following cone courses. Even in the very early days of this technology, the DARPA Grand Challenge test I watched in 2005 included some randomly placed barrels that the vehicle had to maneuver around. You might make a case that following a cone course, at the limit of adhesion, does make the FSD competition technically interesting, but without differing surfaces (ie, wet), there isn't much application to emergency evasion on real roads.

Looking at this press release, http://www.egr.msu.edu/news/2017/04/05/autodrive-challenge the reason for the three year development becomes clear, it lays out goals in terms of robust control in an unpredictable environment--
> Year 1 will focus on concept selection for university teams by having them become familiar with sensing and computation software. They will be tasked with completion of a concept design written paper as well as simple missions for on-site evaluation. These simple missions can include straight roadway driving and object avoidance/detection. ...

In other words, Year 1 will already require considerably more "street smart" cars than FSD, due to the requirement for object detection & avoidance.

> In Year 2 the teams will refine their concept selections into solid system developments and will have more challenging dynamic events for testing on-site, including dynamic object detection and multiple lane changing.
> Year 3 will culminate with final validation of design and concept refinement. They will navigate complex objectives of on-site testing, including higher speeds, turnabouts and moving object detection.

[Pat Clarke]

DRAIN THE SWAMP!!! :D

Z

Many here might not understand that reference, but I do and I agree 100% Erik.

The original old saying was "When you are up to your ass in alligators, it's hard to remember the original intention was to drain the swamp"

Think about how that relates to contemporary FS/FSAE competitions!

Pat

[BillCobb]

While you are contemplating self-driving FSAE cars, keep in mind that these vehicles can have very poorly defined, designed and unsigned vehicle dynamics (unlike the Bolts you have mentioned). There already S.D. Bolts out there under control and watched critically.

Therefore the safety aspects of operating such cars in open OR closed areas becomes a nightmare, considering the low test hours, unforeseen maneuvering territories, inexperienced programmers, questionable durability, hacking temptations, space, barriers, crowd control, and the birthpangs of a new pair of dimes.

That means insurance, liability, venue concerns and likely dangerous practices will be pair of mounts.

Say it ain't so. But meanwhile, I'd be willing to bet that Demolition Derby points could be awarded to the victims of Darwin.

Start off with Figure-8 events for race heats. Let the 1st and 2nd place cars then compete for a championship.

[Swiftus]

@Bill

I think everything you said could be applied to the current state of Formula Student events around the world with a single and possibly critical distinction. An event running a formula driverless competition has the option of installing their own event-designed remote kill switch to the self driving vehicles. As of right now, the 'driver-full' cars are capable of having an inexperienced and panic-prone student at the helm, ready and able to ignore all signaling from officials or car control logic at their own whim.

The idea of a poorly-built FS car careening out of control is as old as FS itself. I'm not sure that argument can hold muster any better than its 'driver-full' counterpart with respect to all of the insurance / liability / venue concerns etc. that were raised.

Personally, I can see the future of Formula Driverless being a bit more exciting than Formula Student since it would allow for a 'safe' method of wheel to wheel driving. For example, run all of the standard Formula Student events and then allow the 'Top 5' runoff to be a wheel-to-wheel race. This kind of additional feature would add a great set of the additional problems which are being faced by road-going vehicles. While it may not be true that a pedestrian would walk in front of a race car, collision avoidance from another car or a set of cars in front would provide a very similar problem to solve to the car's AI.

[Will M]

Seems to me autonomous fsae would absolutely require wheel to wheel racing.
The challenge is navigating a dynamic area with other cars, not just following a line.

The best part about it would be how 'daring' the cars could be.
The cars could race at 10/10ths and if they crash you don't need marshals to save the driver, just let the car sit (if it is off track).
I would envision a track with zero humans in the field and everyone at a safe distance.

The simplest way to achieve this would be with a spec go-kart chassis, engine, drivetrain, and brakes.
The aero and autonomous-ness would be where they competed.
The competition would not be held along side regular FSAE events but rather at tracks where go-karts are already raced.

By focusing on specifically on the autonomous-ness and using a spec go-kart the project complexity is reduced and a one year design cycle is more readily achieved.
The aerodynamics would be open for competition because they would depend greatly on what sensors are used.

-William

[Will M]

I'm confused, it has been days...
Why are you all not telling me how bad my idea is?
This is not like you; should I call the doctor?

-William

[theTTshark]

It was such a bad idea everyone but me has quit the forum. Well done, you win...... ;)

[JulianH]

Will,

fine. I'll bite.

1) FSAE should stay FSAE. If you want to play with stock Go-Karts; that's another racing series.
2) >80% of all teams are already struggeling to put a car together that can run 22km. Putting them wheel to wheel is prone to massive failure and big cost.
3) Driverless FSAE is just one step more as regular FSAE. In the "perfect world" (let's say 3-4-5 years down the road) we will have only driverless FSAE cars that are perfectly designed to the new challenge and not "re-modeled" regular FSAE cars.
Then the challenge stays the same as in FSAE right now, just with one more feature that is to be designed, tested and improved. The "design challenge" stays the same.
Just that we don't have to recruit the best drivers, we have to create them ourselves.

[Will M]

JullianH,

I do agree that it should be a totally separate event under the collegiate design series umbrella, and I should not have called it autonomous fsae.
I propose using existing platforms ( ie go-karts) specifically to reduce the scope of the event and to let team focus on their key autonomous systems.

I would argue against teams build their own platforms for an automous series.
Like you said most team struggle to complete endurance and adding autonomous systems greatly increases the challenge.
And while interdisciplinary projects are great I feel the cost of such additonal complexity out weighs the added learning opportunity.

-William

[theTTshark]

Then the challenge stays the same as in FSAE right now, just with one more feature that is to be designed, tested and improved. The "design challenge" stays the same.
Just that we don't have to recruit the best drivers, we have to create them ourselves.

I think people underestimate how much you can engineer a driver as well....Either way, I think it's easier to find a driver who will get 95% out of the car then it is to design an autonomous system to do that same thing. Personally, I have no interest in an autonomous FSAE because part of the design challenge is that a human driver doesn't always make sense. They don't always make the right decisions. They don't feel or think through things perfectly. And just like the rest of FSAE, this is an engineering management competition first and an engineering design competition second if we're honest about it. It's good to learn how to manage the "end user" as well I think.

[noah]

Announced today via twitter (and yes, I can already hear the groans):
https://pbs.twimg.com/media/DFcXagPXcAAoDmZ.jpg
https://twitter.com/FormulaStudent/status/889205210727559170
(https://twitter.com/FormulaStudent/status/889205210727559170)
Unclear if it will be based on Formula Student Vehicles or not, but given the context of the announcement I would assume so.

[Scott Monash]

Re the FSUK platform, I have no inside info, but I strongly suspect it will use this:

https://www.sbs.ox.ac.uk/programmes/degrees/emba/how-to-apply/news/accelerating-autonomy-emba-alumnus-and-f1-veteran-mark-preston-streetdrone-one

Like many of the other comments I think the new class of Driverless FSAE pioneered by FSG will be hugely challenging and exciting and Monash is certainly interested in joining the fray in future years.

I also agree that it is fine to come up with new autonomous competitions but they are not a direct replacement for FSAE.

All going well, it would be great to see other major competitions around the world adopt the FSG driverless ruleset, and while they are at it their whole FS ruleset, including the EV voltage limit and the massively reduced complexity rulebook.

They clearly know what they are doing and are investing the time and effort to push this competition forward and refine what we already have.

Scott

[NickFavazzo]

I'm sure that I am not the only engineer that has been able to learn things by driving the car or an older car and applying that experience to their new design decisions. Being able to feel the effect of suspension geometry changes, steering ratios, etc helped me immensely to learn and understand what I was doing. Being able to drive is a huge learning aid.

The auto challenge will be very difficult imo but would certainly be an interesting venture for teams but I don't think it can/should replace 'normal' FS/FSAE.