Meet "Baxter" the Robot Out to Get Your Minimum-Wage

[jneal3]

Re: technology policy vs labour

http://news.drive.com.au/drive/motor...115-2crgk.html

Automated (technically as opposed to autonomous) driving is coming more quickly than people realize. A major insurance company has been giving out research grants to automotive OEMs to give them intimate access to the development of this technology. It's time to think about how this will change things, and where the opportunities for gain lie.

[c1ue]

Re: technology policy vs labour

Automated (technically as opposed to autonomous) driving is coming more quickly than people realize. A major insurance company has been giving out research grants to automotive OEMs to give them intimate access to the development of this technology. It's time to think about how this will change things, and where the opportunities for gain lie.

You might look at the post earlier where I talk about the problems associated with automated parking.

There is technology to parallel park - Toyota has had that for several years now.

However, parallel parking, while a challenge to those who don't have to do this often, is trivial compared to the task of finding a parking spot.

It is estimated that 30% of all traffic congestion in San Francisco is due to people looking for parking. If you drive along Fell street between downtown/freeway exit on Octavia and Golden Gate Park - which is a major east/west corridor for those who live/know the City, driving around 5 pm is a huge hazard because of those who live along there trying to parallel park in between the brief intervals where the synchronized traffic lights permit - i.e. racing like mad to get in position, slamming on the brakes and starting to back up before traffic fills in all the lanes between the red lights. All you need - if you're trying to drive past that area - is for one car on the right and one on the left to try to park, then you get to see 3 lanes of cars attempt to squeeze into one lane in the span of one block.

A lovely test scenario for the automated car.

[wayiwalk]

Re: technology policy vs labour

You might look at the post earlier where I talk about the problems associated with automated parking.

There is technology to parallel park - Toyota has had that for several years now.

However, parallel parking, while a challenge to those who don't have to do this often, is trivial compared to the task of finding a parking spot.

It is estimated that 30% of all traffic congestion in San Francisco is due to people looking for parking. If you drive along Fell street between downtown/freeway exit on Octavia and Golden Gate Park - which is a major east/west corridor for those who live/know the City, driving around 5 pm is a huge hazard because of those who live along there trying to parallel park in between the brief intervals where the synchronized traffic lights permit - i.e. racing like mad to get in position, slamming on the brakes and starting to back up before traffic fills in all the lanes between the red lights. All you need - if you're trying to drive past that area - is for one car on the right and one on the left to try to park, then you get to see 3 lanes of cars attempt to squeeze into one lane in the span of one block.

A lovely test scenario for the automated car.

That is a great example of a "real problem" instead of the defined one that isn't a problem at all to solve.

I think there will be other interesting outcomes (parking on top of broken bottles/glass, in the middle of large puddles, etc).

Should be fun!

[Polish_Silver]

Re: technology policy vs labour

http://news.drive.com.au/drive/motor...115-2crgk.html

Automated (technically as opposed to autonomous) driving is coming more quickly than people realize. A major insurance company has been giving out research grants to automotive OEMs to give them intimate access to the development of this technology. It's time to think about how this will change things, and where the opportunities for gain lie.

I believe we will start parking, queuing up in stop and go situations and then next step will be autobahns,” he said. “It will be a move into 100 per cent autonomous driving.”

Autonomous car on the autobahn should be easier than city streets. The controlled access road means that you don't have children or pets running into the street, people pulling out of driveways, bicycles, etc. The visual field is greatly simplified.

[DSpencer]

Re: technology policy vs labour

You might look at the post earlier where I talk about the problems associated with automated parking.

There is technology to parallel park - Toyota has had that for several years now.

However, parallel parking, while a challenge to those who don't have to do this often, is trivial compared to the task of finding a parking spot.

It is estimated that 30% of all traffic congestion in San Francisco is due to people looking for parking. If you drive along Fell street between downtown/freeway exit on Octavia and Golden Gate Park - which is a major east/west corridor for those who live/know the City, driving around 5 pm is a huge hazard because of those who live along there trying to parallel park in between the brief intervals where the synchronized traffic lights permit - i.e. racing like mad to get in position, slamming on the brakes and starting to back up before traffic fills in all the lanes between the red lights. All you need - if you're trying to drive past that area - is for one car on the right and one on the left to try to park, then you get to see 3 lanes of cars attempt to squeeze into one lane in the span of one block.

A lovely test scenario for the automated car.

Sounds like we should build an automated parking garage.

[c1ue]

Re: technology policy vs labour

Autonomous car on the autobahn should be easier than city streets. The controlled access road means that you don't have children or pets running into the street, people pulling out of driveways, bicycles, etc. The visual field is greatly simplified.

The autobahn has less types of interference, but the interference which does occur is much, much higher speed (i.e. lower response time).

Cue any number of idiotic driver Youtube videos, or Dukes of Hazzard reruns.

Sounds like we should build an automated parking garage.

A quasi-form of this exists in extremely high land cost areas like Tokyo: essentially cars in a few specialty parking garages are put into an Amazon type bucket, but are then stored in a 3D array (Amazon also uses 3D arrays, but pulls only from the top).

The problem is that this equipment is extraordinarily expensive. Another problem is that 2/3rds of all parking spots in a typical city is on the street, and of the 1/3 which are parking lots/garages, a large fraction are pure 'free rent' holding operations (i.e. the parking lot is offering parking only until the lot is sold off to make a building). These operations have even less incentive to invest.

[mmr]

Re: technology policy vs labour

It's already been tried - the first version didn't work out so well...

http://www.nj.com/hobokennow/index.s...t_hoboken.html

[c1ue]

Re: technology policy vs labour

An interesting note in this article about the self driving car:

http://pandodaily.com/2013/03/08/all...f-driving-one/

Lidar technology – integral to the self-driving car’s very functionality, essentially serving as the car’s eyes – is right now too vulnerable to the forces of nature: snow, dust, even glaring sunlight.

This would seem to be a pretty major issue.

[jneal3]

Re: technology policy vs labour

An interesting note in this article about the self driving car:

http://pandodaily.com/2013/03/08/all...f-driving-one/



This would seem to be a pretty major issue.

Yes, that and cost are major issues. Research (http://web.mit.edu/newsoffice/2013/c...ight-0109.html) is progressing, but I think the reason Daimler went with stereo cameras is that lidar (the 360 degree kind) is not commercially feasible.

The discussion of self-driving cars gets mixed up between 'automated' and 'autonomous'. Autonomous driving is not coming in 3-5 years. Automated driving will become common fairly quickly, with the payoff being a potentially large reduction in the 33,000 deaths/year, 93% of which are attributable to human error (btw I'm involved in pedestrian impact avoidance testing, the recognition algorithms have come a long way since the Volvo tests EJ cited, and they're improving at a remarkable rate, and Volvo is absolutely correct that dummy testing is not a good surrogate for how humans are perceived by the recognition systems). I'm pretty sure Google is wildly naive thinking that they're going to give blind people the ability to drive in 3 years:

http://www.youtube.com/watch?v=TJGsb5XgU2k

[DSpencer]

Re: technology policy vs labour

A quasi-form of this exists in extremely high land cost areas like Tokyo: essentially cars in a few specialty parking garages are put into an Amazon type bucket, but are then stored in a 3D array (Amazon also uses 3D arrays, but pulls only from the top).

The problem is that this equipment is extraordinarily expensive. Another problem is that 2/3rds of all parking spots in a typical city is on the street, and of the 1/3 which are parking lots/garages, a large fraction are pure 'free rent' holding operations (i.e. the parking lot is offering parking only until the lot is sold off to make a building). These operations have even less incentive to invest.

Yeah I've seen videos of the ones in Japan. How expensive are they? I'm not quite sure I follow your line of thought on the existing parking situation. Are you saying that there is too much competition from cheap parking already? It seems like if people are swarming over the rare street spot that at least some would be willing to pay for a more convenient and secure spot. I know I would pay a small amount to avoid searching for a spot and pulling off a rapid parallel parking job with a line of traffic behind me. I don't know the economics though. I wonder how many cars and cost per car it takes to make money on the large upfront investment.

Also mmr's link seems to suggest the technology leaves something to be desired. Or at least did a few years ago.

[c1ue]

Re: technology policy vs labour

Yeah I've seen videos of the ones in Japan. How expensive are they?

I don't have direct access to the numbers on the 3D lots in Japan, but many lots in Tokyo have a 1.5D setup (i.e. individual spaces can hold 2 or 3 cars). The cost of these vs. a structure is in the order of 2.5x, and a structure's cost vs. just a paved lot with stripes painted is 10x. Labor costs are, however, lower as either the 1+D or structure's capacity increases.

I'm not quite sure I follow your line of thought on the existing parking situation. Are you saying that there is too much competition from cheap parking already? It seems like if people are swarming over the rare street spot that at least some would be willing to pay for a more convenient and secure spot. I know I would pay a small amount to avoid searching for a spot and pulling off a rapid parallel parking job with a line of traffic behind me. I don't know the economics though. I wonder how many cars and cost per car it takes to make money on the large upfront investment.

My point about many of the lots being just economic free rent holding patterns is that parking is simply not profitable enough to justify economic investment beyond the paved/painted stripe level. Ultimately no storage area for cars can compete with a multi-story commercial building, and holding costs are so low (and mitigatable via selling parking) that few bother.

As for the economics of parking: the problem is predictability. Parking operators want guaranteed income - they are completely uninterested in the ebb & flow business and focus all their efforts on their monthly customers (acquisition and retention). In fact most parking operators I speak to have only a tiny fraction of their inventory even available to 'drive bys'.

On the street, the problem is there is no way to guarantee access even if parking lot rates were charged. The dynamic pricing is an attempt to resolve this, but the ultimate effect is to just drive up the cost of parking overall - already in SF there is a dynamic of increasing street parking costs feeding into higher parking operator charges.

Unfortunately the ultimate effect is to just drive up the cost of living. SFPark, the program which installed all of the 'smart' meters in SF - which in turn was largely paid for by the federal government - added 10000 meters to SF. This increased the overall meter supply from about 30K to 40K, or 2.8% more of the street parking is now metered. In concrete terms, this has translated to large swaths of the city no longer having any type of free parking available, which in turn has pushed all the residents in those areas to park further and further from where they used to. Which in turn ripples outward.

So to summarize: even beyond the problems with GPS accuracy (for any smart phone/app based solution), smart parking in a large city not only has a data problem (policy data isn't generally usable form - which is what I do), but also a pricing, customer education, and operational problems.

Pricing because the private operators want predictability. From their own experience, they know that it is far better to get a monthly customer in the hand vs. dozens of 'drive bys' in the bush. For those cases where they know demand is going to skyrocket over supply, they in turn will charge exorbitant rates - 5x or even 20x normal. In this case, it is worth the risk of having unsold inventory.

For the cities, getting better revenue from meters also dovetails nicely with making sure merchants' customers have parking. However, the dynamic pricing systems which are intended to accomplish this have already shown a severe problem: it is now impossible to understand what the price of parking is at a meter in SF. In the past, the range varied from $2/hr to $3 to $3.50 in distinct zones. Now, the meters can vary from $0.25 to $5.50, with $10/hr or $18/hr for events, with no way for you to know barring use of an app while driving.

Customer education is a problem: everyone wants cheap, close, and easy parking. Unfortunately it is literally impossible to achieve all 3 simultaneously when you have 380K registered vehicles in SF plus another 200K to 300k visitors (commuters, tourists etc) vs. 350K on and off street parking spaces (i.e. public and commercial, not including private). Customers willing to park further away and/or pay more, that would be one way to resolve this dilemma, but they aren't willing now - mostly.

Operational problems - in that there simply aren't mechanisms to operate smart parking even if the above 2 areas are resolved. If most parking operators aren't even willing to install remote access and payment systems (which I can attest to personally), aren't willing to have their personnel even update via text/web page/phone call (also first hand experience), then the problems of preserving street parking or even increasing parking operator inventory for drive bys just aren't going to be resolved anytime soon.

[radon]

Re: Robots are replacing humans

You still don't get it.

Sure, fry cooks are interchangeable from a certain perspective.

So are employers.

However, people aren't inanimate objects. If you treat them as such, then you're going to reap what you sow.

As for ad hominem - what I've repeatedly said is that I'd never want to work for someone who thinks of employees like you do. If you treated your wife, your kids, your friends the same way, you'd equally get the same result.



You don't seem to actually know anything about robotics and automation - instead merely focusing on what you think you'll avoid by not having those uppity people working.



As I am not the only one speaking to the fundamental limitations robots and automation have - your comment once again is completely off the mark.

I noted that automated delivery trucks can't actually deliver. EJ noted the same, although he added that maybe customers can be made to take this over.

I noted that automated driving for buses incurs fundamental limitations because bus drivers do more than just move from point A to point B. EJ noted that auto-driving in general has limitations because robot vision is only pattern match based.

All the other points raised - they didn't arise from thin air. I either know people who do this for a living or have direct hands-on experience with technology ranging from semiconductors, to computer design, to mobile apps.

But hey, you clearly don't want to know that your dream robot employee is, right now and for the near future, a fantasy.

No, I certainly do get it. I've spent enough time implementing SVMs and the like to have a pretty good idea of the challenges involved in pattern recognition and adaptive behavior. I've also spent my fair share of time developing mobile applications, and I'm certainly aware of the power issues involved and have personally experienced their impact on the development cycle, especially when third party drivers are involved as they are in most commerce applications.

Furthermore, I know enough electrical engineers not to take their word on everything, especially on something that is rapidly becoming a software problem. Since I don't know you personally, it boils down to whether I accept what a random person on the internet said to be true. In this case you have given me no reason to convert to your opinion. If you could provide more concrete examples I would be more inclined to weigh your position, but none are forthcoming. For instance, if your opinions were as well supported as EJs, we wouldn't be having this discussion.

When I posted this story about Baxter last September it was because I thought it was an interesting idea and I wanted to see what people thought about it and the possible economic fallout of cheap robotic labor. I didn't care whether it actually worked and still don't. Early adopters nearly always get burned. But saying that building a robot like that is impossible is accurate, even if this one fails to work as advertised. Just because I can't think of a way to accomplish something doesn't mean that it is impossible or that some smart guy with resources a Google isn't doing it. It is the reason I solicit options here, there would be no point otherwise.

[reggie]

Re: Robots are replacing humans

Forget about automated parking, My fav is Garnet Hertz's Roach Robot they've been working on at Conceptlab, UC Irvine...

Cockroach Controlled Mobile Robot" is an experimental robotic system that translates the bodily movements of a living, organic insect into the physical locomotion of a three-wheeled robot. Distance sensors at the front of the robot also provide navigation feedback to the cockroach, striving to create a pseudo-intelligent system with the cockroach as the CPU.

http://www.conceptlab.com/roachbot/

[metalman]

Re: Robots are replacing humans

Forget about automated parking, My fav is Garnet Hertz's Roach Robot they've been working on at Conceptlab, UC Irvine...

Cockroach Controlled Mobile Robot" is an experimental robotic system that translates the bodily movements of a living, organic insect into the physical locomotion of a three-wheeled robot. Distance sensors at the front of the robot also provide navigation feedback to the cockroach, striving to create a pseudo-intelligent system with the cockroach as the CPU.

http://www.conceptlab.com/roachbot/

kids these days... cockroach crawling on a computer mouse ball... randomly driving a robot... zzzzzzzzzz.

Distance sensors at the front of the robot also provide navigation feedback to the cockroach

did the cockroach read the manual?

[flintlock]

Re: technology policy vs labour

That is a great example of a "real problem" instead of the defined one that isn't a problem at all to solve.

I think there will be other interesting outcomes (parking on top of broken bottles/glass, in the middle of large puddles, etc).

Should be fun!

I call some of this stuff "technology in search of a problem". Is parking a car really that high on most people's list of problems? Driving one? I like some of the accident avoidance technology, especially the one that keeps you from rear ending the car in front of you. But No, the actual reason for a lot of this stuff being dreamed up is to find a way for some tech firms to continue to exist, as quite a bit of technology today is pure fluff and they know it. Then they turn it over to the marketing people to convince us we "need" it.