Cadillac Super Cruise almost here - Motortrend drives it

[calisnow]

Okay Elon - you've got 4-8 weeks to get another "even silkier" build of AP2 on the roads before these Caddy's are released into the wild and someone could finally argue a plausible claim that there is a smoother, more functional AP system available for highways.

Overall positive article on GM's hands free system went online today: GM's Super Cruise: Driving a 2018 Cadillac CT6 With a Full-On Auto Pilot - Motor Trend - competition is excellent. I imagine car forums over the next few years are going to be filled with endless "My autopilot is better than yours" thread wars - this is only the beginning...

"Sampling the system in an early-build 2018 CT6 Platinum, it functioned in a decisive way that instantly inspired confidence. Upon entering the freeway it quickly acquired the center of the lane and stuck to it faithfully, slowing quite gently when encountering slower-moving traffic then resuming the set speed in a timely fashion following a lane change. The infrared emitters and camera were able to penetrate my mirrored Oakleys with ease, and I was never hectored to pay attention or resume looking at the road except when I truly looked well away for extended times. There were no false alarms, in other words. One Tesla feature Cadillac is missing is the display indicating the actual path of the upcoming roadway with icons indicating car or truck traffic in the nearby lanes. The only display option is the typical adaptive cruise one with straight lane markers and a Cadillac CT6 rear end displayed when a vehicle is sensed ahead."

• "Highways only. The system only becomes available once you’ve entered a meticulously lidar-mapped, divided, limited-access highway in the U.S. or Canada—or a limited-access stretch of a highway that switches between on-ramps and crossings (like California Highway 101). When the steering wheel icon appears at the upper right of the central speedometer gauge, press the Super Cruise button, and when the car has locked on to the center of the lane, the light bar on the top of the steering wheel turns green.

• Pay attention. Infrared emitters on the top quadrant of the steering wheel illuminate the driver’s head and face, and a steering column–mounted camera with infrared capability (for night detection) constantly monitors head position and eye focal point. Look away for 15 seconds or less, and the system demands you return your gaze to the road by first flashing the green light, then flashing red, sounding a tone, vibrating the seat, and ultimately issuing a voice warning. If none of that succeeds in returning the driver to the task of controlling the vehicle, it will slow to a stop in the lane of travel, put the hazard flashers on, and summon help via OnStar. If you get to those last stages and resume control, Super Cruise locks you out until the next time you stop and restart the car. A nice touch: capacitive sensors detect the driver’s hand(s) on the wheel, eliminating the need to wiggle it to verify control as with some lane keep assist vehicles.

• You change lanes. Any deviation from the absolute center of the marked lane is on you. Want to avoid a pothole? That’s your job. Need to change lanes to pass a slowpoke? Check your blind spot (the car will monitor it, too), signal, and steer into the desired lane (overcoming some resistance torque initially). The light bar on the wheel turns blue while you’re driving. Center the car in the new lane and wait for the steering wheel bar to return to green, signaling Super Cruise is back in charge.

• Reasonable weather. Autonomy is all about sensor fusion and multiple redundancies. Here the system relies on forward- and side-looking cameras, long-range radar, high-precision GPS (courtesy of Trimble), and a GM-proprietary high-def map database generated in conjunction with and maintained by a tech startup GM Ventures invested in called Geodigital). If the camera or radar data becomes compromised by heavy rain or snow covering lane markings or obscuring the radar unit or if the GPS system somehow goes down, the system will call upon the driver to resume control. Until that happens, the system will continue steering (or come to a stop as noted above), relying on all sensors that still work. Tunnels deprive the system of GPS, but maps keep it working for up to 1 km (0.6 mile), after which the driver is asked to take over.

• Construction zones. The owners’ manual warns that the system is not programmed to cope with construction zones, but in practice it handles such zones OK if lane markings are provided and travel remains on the main road surface. But for example it will not follow even a well-marked deviation that crosses the median to share the opposite direction’s roadway.

A few more interesting features:

• Curve Speed Control. When Super Cruising, if the radius of a curve on a mapped highway would result in dangerous, uncomfortable, or unnatural lateral g levels, the car will automatically slow for the turn (and the cruise-control icon at the upper left of the main speedometer display indicates this system is active).

• Regular and Adaptive Cruise. Whenever you’re not on a Super Cruise–qualifying highway, you now have the choice of adaptive or ordinary cruise control. (Press and hold the cancel switch to toggle between these modes.) This feature is spreading across the whole Cadillac lineup for 2018. Oh, and the time-out length after which the system won’t automatically resume driving in a stopped traffic jam is stretched from a few seconds to more like half a minute when in Super Cruise mode.

• Pingpong lane keeping. In you-drive mode the Super Cruise sensors are still informing all other safety systems, such as the pre-collision warning and braking, lane departure warning, and lane keep assist. The latter still lets you wander from lane marker to lane marker because centering in the lane could mislead drivers into thinking Super Cruise was engaged."

 

[ItsNotAboutTheMoney]

Sounds good.
Hopefully it's priced to sell and it won't be too long before it moves down to Buick and Chevrolet.
Also, I hope that the arrival of Supercruise will make GM treat regular adaptive cruise control as a mainstream feature and it'll be added to other models.

 

[McRat]

Sounds good.
Hopefully it's priced to sell and it won't be too long before it moves down to Buick and Chevrolet.
Also, I hope that the arrival of Supercruise will make GM treat regular adaptive cruise control as a mainstream feature and it'll be added to other models.

ACC is available on a number of GM cars, and it's pretty competent. I personally don't like it in EV driving because it harms the range.
GM ACC hunts gears in the mountains with automatic transmissions, which can be annoying. But this is nothing new to ICE cruise controls in steep mountains. To get around this, switch to manual shift mode and hold a gear, IIRC, it's about 5th or 6th for most freeways through mountains. It's one more reason EV powertrains are superior.

But on level ground in medium to heavy traffic it's a pleasure to use. Very smooth, but can also nail the brakes when someone begins to cut you off dangerously to safely avoid a collision. When somebody cuts you off with 1 sec or more of following distance, it gently adjusts speed. Where it wastes energy, is when somebody changes lanes from in front of you, then the road is wide open. It will use excess acceleration IMO to catch up to the traffic ahead. This can be attenuated by running the following distance at 3 seconds, which smooths this out some.

Interesting tidbit: The "Build Your Own" 2018 CT6 webpage shut off yesterday and is still off today. It used to be working, but with no mention of Super Cruise for the 2018's which are already on the road.

It will be interesting to see the price of a Super Cruise Caddy. Keep in mind the MSRP of the Platinum AWD 3.0TT all-wheel steering CT6 is a heady $90k, but right now great deals can be found. Also note that there is a 30 mile AER PHEV version of the car now. Will it get the Super Cruise? Stay tuned. The PHEV has issues though. It doesn't have the instant punch the Volt and ELR have. Sometimes. You can't easily predict when turbo lag is going to kick your butt since the PHEV CT6 is the 'classic' PiP style of PHEV, where most the power is ICE, and full acceleration requires the ICE to be warmed up and running.

Note that Cadillac was the first GM to actually use an Over-The-Air update for a driving function in 2013 in the ATS. Prior to that OTA was for OnStar related systems only. It is entirely possible the Super Cruise system will be enhanced via OTA. Most GMs have 4G connectivity and OnStar available for bi-directional data transfer since 2006? 2009? I forget. 2004 you had to bring it in for OnStar updates.

 

[electracity]

Very interesting. Is Supercruise Mobileye's next version? GM makes a claim of proprietary maps, just like Tesla did with AP1. The formula: Implement someone else's technology, don't mention that vendor, make a proprietary claim to help mask who owns the IP.

Tesla got on the path of camera based mapping through fleet learning from mobileye. There is ample evidence that mobileye has abandoned that strategy. The chance that Tesla is on a technological dead end is increasing.

 

[calisnow]

Interesting tidbit: The "Build Your Own" 2018 CT6 webpage shut off yesterday and is still off today. It used to be working, but with no mention of Super Cruise for the 2018's which are already on the road.

They're on the road already? Thought they hadn't been built yet.

@electracity - Super Cruise is some kind of homologation tech. GM was working on their own for several years - then they purchased for around a billion dollars in cash and stock a 12 man startup run by an MIT dropout 2 years ago called Cruise Automation. CA was originally going to release a self steering aftermarket add-on for the Audi A4 - and took deposits - but then apparently impressed GM execs so much in a private demo they simply swallowed the company.

 

[calisnow]

Driven In The Valley: The Startup Founders Fueling GM’s Future

GM lets its autonomous unit be autonomous - "

DETROIT — Despite spending more than $1 billion to acquire and expand Cruise Automation, General Motors is keeping its distance — both figuratively and literally — from the autonomous-vehicle startup.

The two companies operate in different spheres: GM, rooted in Detroit and drawing on its century of manufacturing experience, focuses on the production of the test vehicles and integration of the self-driving hardware, while Cruise, based in San Francisco, develops and refines the software that controls the systems.

The unique relationship was described by GM CEO Mary Barra and Doug Parks, head of GM autonomous technology and vehicle execution, during an event last week to mark the production of 130 second-generation, self-driving Chevy Bolt EVsat GM's Orion Township, Mich., factory.

"Cruise Automation is running as a startup," Barra said. "Not only are they responsible for the technology, but they're responsible for the commercialization — so the entire business."

For GM, that means resisting the temptation shared by many automakers historically to control and fully integrate the companies they acquire. Company executives say it's a needed change as the legacy auto industry learns to work more constructively with Silicon Valley technology partners. And it spares the two companies the time and disruption of integrating two very different corporate cultures."

 

[electracity]

@electracity - Super Cruise is some kind of homologation tech. GM was working on their own for several years - then they purchased for around a billion dollars in cash and stock a 12 man startup run by an MIT dropout 2 years ago called Cruise Automation. CA was originally going to release a self steering aftermarket add-on for the Audi A4 - and took deposits - but then apparently impressed GM execs so much in a private demo they simply swallowed the company.

GM Reveals More About Supercruise | TheDetroitBureau.com

 

[scottf200]

It compares the data from its own camera and radar with information contained in a highly detailed 3-D map. That’s far more accurate than the current, satellite GPS-based navigation systems of today, but it requires extensive new mapping of roads around the country using a laser-based system called LIDAR, or Light Detection and Ranging.

"Highways only. The system only becomes available once you’ve entered a meticulously lidar-mapped, divided, limited-access highway in the U.S. or Canada

This is a way different system [than Tesla's that does auto-pilot on the fly for the most part] if it has to rely on previous meticulously lidar-mapped roads!

 

[pakman00]

Curve Speed Control...

Tesla programmers, hope you're paying attention to that...

 

[electracity]

This is a way different system if it has to rely on previous meticulously lidar-mapped roads!

If you are saying Cruise Automation running the Bolt is a different effort than SuperCruise on a Cadillac, I agree.

Smart manufacturers are managing multiple technological efforts. Tesla's failsafe is probably Google. It would be nuts for GM to turn over the fate of a large corporation to the couple of nerds who started Cruise Automation.

 

[scottf200]

This is a way different system if it has to rely on previous meticulously lidar-mapped roads!

If you are saying Cruise Automation running the Bolt is a different effort than SuperCruise on a Cadillac, I agree.

Smart manufacturers are managing multiple technological efforts. Tesla's failsafe is probably Google. It would be nuts for GM to turn over the fate of a large corporation to the couple of nerds who started Cruise Automation.

Way different than Tesla's which mainly does autopilot on the fly vs having to rely on *LIDAR* mapped roads.

High quality and accurate LIDAR is expensive. See: Cheap lidar sensors are going to keep self-driving cars in the slow lane

 

[Saghost]

Interesting. I still think allowing ACC and blind CC on the same car is a mistake, and it sounds like Cadillac is making it a simple toggle on the steering wheel - seems like an accident waiting to happen.

The steering wheel lights and the two operating modes (center of lane or hand steer with Supercruise still enabled) are an interesting addition/difference.

I'm disappointed Cadillac didn't copy Tesla's approach of showing what the car is seeing - I feel like that's a big help in knowing when and how much to trust the car.

 

[electracity]

Way different than Tesla's which mainly does autopilot on the fly vs having to rely on *LIDAR* mapped roads.

High quality and accurate LIDAR is expensive. See: Cheap lidar sensors are going to keep self-driving cars in the slow lane

In the case of supercruise, Lidar is used for HD mapping one time by a specialized vehicle. It seems that they have correlated Lidar data with in car camera data. This is probably done with object identification. The car sees a bridge. The car then checks the HD map to find the bridge.

The car compares many objects it sees to the HD map and can know with high probability its location within a very small margin of error.

There is no evidence that Tesla fleet learning producing HD mapping is working. If you read mobileye's white papers it is clear that fleet learning using low bandwidth is a mobileye innovation. If mobileye has abandoned this approach Tesla may be on a different now path too.

The article you linked doesn't mention that google is building their own solid state LIDAR.

 

[Saghost]

Curve Speed Control...

Tesla programmers, hope you're paying attention to that...

TACC/AP does this. I'm not always sure it has the perfect settings, but it will slow down for steeper curves with no other cars around.

 

[electracity]

TACC/AP does this. I'm not always sure it has the perfect settings, but it will slow down for steeper curves with no other cars around.

If/when Tesla runs off HD maps the curves will be easily handled. The lateral gforce of a road sectionis determine in advance. The user could even set their comfort level for g forces.

Tesla today is mostly doing lane keeping and vehicle awareness. Pretty soon that level of functionality will look crude.

Tesla knows all this, of course. They know they have to be able to position the vehicle precisely and know the road ahead. I just don't think customer cars will do the initial mapping.

 

[chillaban]

@verygreen did mention that the 2017.28 firmware had more hints of a codename "atlas" mapping system for AP2. And that AP2 currently does not use any mapping info at all for road features.

It's very much conceivable that AP2 will take a huge leap forward when that functionality becomes activated, and one can argue that Tesla's approach of relying on fleet-learning is more scalable than GM (e.g. their partner's) rather costly and labor-intensive approach of driving around with LIDAR and mapping the roads the system is meant to work on.

But at any rate, love seeing some competition here in the driver assist field. It'll be a very interesting few months ahead.

 

[RubberToe]

Is there a map of the entire U.S. showing which roads the Supercruise feature works on?

RT

 

[McRat]

I think the main feature that will delight today's drivers is that you will probably be able to text 2 handed as long as glance up when the alert comes on now and then.

That's what we really need. A way to keep texters from trying to kill the rest of us.

 

[electracity]

I think the main feature that will delight today's drivers is that you will probably be able to text 2 handed as long as glance up when the alert comes on now and then.

That's what we really need. A way to keep texters from trying to kill the rest of us.

That would be great. My daughters might survive to provide me with grandchildren. Phone too. Eyes on the road, brain on the phone call.

 

[electracity]

@verygreen did mention that the 2017.28 firmware had more hints of a codename "atlas" mapping system for AP2. And that AP2 currently does not use any mapping info at all for road features.

It's very much conceivable that AP2 will take a huge leap forward when that functionality becomes activated, and one can argue that Tesla's approach of relying on fleet-learning is more scalable than GM (e.g. their partner's) rather costly and labor-intensive approach of driving around with LIDAR and mapping the roads the system is meant to work on.

But at any rate, love seeing some competition here in the driver assist field. It'll be a very interesting few months ahead.

Imagine model 3 demand if Tesla regains its reputation as the leader in autonomous driving features.