Wednesday, April 29, 2009
Tuesday, April 28, 2009
Stereo image alignment check
Test animations contributed by Cristian Orrego Nelson. The vertical displacement here looks bad. However, these images are unrectified, so this could be explained by different lens shapes or centres of distortion.
Ideal image alignment from the Middlebury data set. Notice that there is no vertical movement.
Ideal image alignment from the Middlebury data set. Notice that there is no vertical movement.
Labels:
camera calibration,
stereo vision
Sunday, April 26, 2009
Sentience project microblog
This is probably rather self-indulgant, since I think I'll be blogging to an audience of zero, but I've set up a microblog for the Sentience project on floss.pro, which I'll use to occasionally comment on progress with GROK2.

Anyone observing the home pages of the Sentience and Robobridgeware projects could be forgiven for believing that very little progress has occurred there for a year or more, but this could hardly be further from the truth. Most of the stuff which I've been doing has been unglamourous work on the low level systems, getting one system to talk to another, simulation/unit testing and bug hunting - and there have been some most heinous bugs. Progress has been slow, since this is only a hobby project which I work on maybe one day per week, but improvements have been accumulating and eventually a critical mass of software will be achieved.
The GROK2 robot, which I'll be using for tests, is only intended to be - to borrow the phrase used by Donald Michie - a "research purpose robot". I'm not intending it to be of any commercial value or interest. However, I have been trying to write the software in such a way that there's a strong degree of hardware abstraction, which should allow me to easily tailor the same system to other types of robot in the future, with minimal rewriting of code. The software design is also very friendly to multi-core systems, so that as more processing resources become available performance will increase.
In the longer term I expect that consumer priced PC based robots will become a reality, after countless years of struggle and false dawns. Once that happens commercial applications might become something worth investigating. If I get to the stage where I think I have visual perception software which could be substantially useful, and am sufficiently frustrated by lack of progress on the hardware side of things, I might even try to start a PC-bot business myself, since I think I have most of the requisite skills (except for sales/marketing).

Anyone observing the home pages of the Sentience and Robobridgeware projects could be forgiven for believing that very little progress has occurred there for a year or more, but this could hardly be further from the truth. Most of the stuff which I've been doing has been unglamourous work on the low level systems, getting one system to talk to another, simulation/unit testing and bug hunting - and there have been some most heinous bugs. Progress has been slow, since this is only a hobby project which I work on maybe one day per week, but improvements have been accumulating and eventually a critical mass of software will be achieved.
The GROK2 robot, which I'll be using for tests, is only intended to be - to borrow the phrase used by Donald Michie - a "research purpose robot". I'm not intending it to be of any commercial value or interest. However, I have been trying to write the software in such a way that there's a strong degree of hardware abstraction, which should allow me to easily tailor the same system to other types of robot in the future, with minimal rewriting of code. The software design is also very friendly to multi-core systems, so that as more processing resources become available performance will increase.
In the longer term I expect that consumer priced PC based robots will become a reality, after countless years of struggle and false dawns. Once that happens commercial applications might become something worth investigating. If I get to the stage where I think I have visual perception software which could be substantially useful, and am sufficiently frustrated by lack of progress on the hardware side of things, I might even try to start a PC-bot business myself, since I think I have most of the requisite skills (except for sales/marketing).
Labels:
GROK2,
microblogging,
sentience
Saturday, April 25, 2009
Friday, April 24, 2009
Cyberdogs
Part dog, part video recorder. This could be very useful for search and rescue type missions, for example after an earthquake. It's also a demonstration of how small and light weight camera and storage devices have become.

I wonder whether you could do monocular SLAM from a dog mounted camera. Cameras are so small now (for example the ones used in mobile phones) that multiple cameras could be used on the device to produce a real time composite 360 degree view around the dog - the doggy egosphere. The possibilities for dog related gadgets are endless.

I wonder whether you could do monocular SLAM from a dog mounted camera. Cameras are so small now (for example the ones used in mobile phones) that multiple cameras could be used on the device to produce a real time composite 360 degree view around the dog - the doggy egosphere. The possibilities for dog related gadgets are endless.
Spooks and greenies
When I think of spies I inevitably think of James Bond, gadgets, Bond girls and jazzy sound tracks, but the job of being a real spook is probably very unglamourous. The Guardian newspaper has some amusing audio of an unnamed spy rather ineptly trying to bribe an environmentalist into divulging information.
Can political actions help to reduce pollution from aviation?
As an engineer I'd say that the scope for political actions to influence the amount of pollution produced, short of totalitarian methods like banning people from going on holiday or taxing them to a ridiculous level, is quite limited. The crux of the problem as I see it is that aircraft engines need to transition from being oil powered to being hydrogen powered. Powering an aircraft on hydrogen makes sense on many levels. Burning hydrogen produces negligible pollution (water vapour) and also would significantly reduce the weight of the aircraft - increasing is passenger carrying efficiency.
So in my opinion environmental groups should be lobbying for R&D money to be invested in new engine technology, with an emphasis on hydrogen power.
"Plane Stupid welcomes actions in its name, provided they are non-violent and accountable and help further the struggle against airport expansion and greenhouse gas emissions from aviation."As far as I know the environmental group involved is a peaceful one, campaigning against the environmental pollution produced by passenger aircraft, so there is probably no need to waste taxpayers money attempting to bribe such people. The police and secret service organisations should instead be concentrating their scarce resources upon groups preaching hatred and endorsing violent tactics - people who actually could cause harm and damage.
"wouldn't it also be nice to have tax-free - money you'd be getting. You wouldn't pay any tax on it. So you could do with it what you want."It's nice to know that government bribes are not tax deductable.
Can political actions help to reduce pollution from aviation?
As an engineer I'd say that the scope for political actions to influence the amount of pollution produced, short of totalitarian methods like banning people from going on holiday or taxing them to a ridiculous level, is quite limited. The crux of the problem as I see it is that aircraft engines need to transition from being oil powered to being hydrogen powered. Powering an aircraft on hydrogen makes sense on many levels. Burning hydrogen produces negligible pollution (water vapour) and also would significantly reduce the weight of the aircraft - increasing is passenger carrying efficiency.
So in my opinion environmental groups should be lobbying for R&D money to be invested in new engine technology, with an emphasis on hydrogen power.
Learning from evolution
I like the simplicity of the hexapod. It looks like it has curved springy legs which are simply rotated - a very simple but effective actuation method.
Raw data: UK national debt
An interesting little nugget of history that I didn't know previously. According to these figures at the time when the second world war began Britain was technically already bankrupt, having a national debt of more than 100% of GDP. It looks like public debt really got out of control after the beginning of the first world war, and didn't recover again until the mid 1970s.
So, although the debts recently announced in the budget are gigantic - unprecedented within the last 30 years - historically Britain has been in far worse debt in the past. After WW2 the UK becomes more than doubly bankrupt, with debt nearing 240% of GDP.
So, although the debts recently announced in the budget are gigantic - unprecedented within the last 30 years - historically Britain has been in far worse debt in the past. After WW2 the UK becomes more than doubly bankrupt, with debt nearing 240% of GDP.
Wednesday, April 22, 2009
Caleb Chung and the making of Pleo
Sadly Ugobe are now bankrupt - another casualty of the recession. What robots like Pleo and the earlier AIBO made by Sony demonstrate is that there is a market for fairly complex robotic toys, and that modern manufacturing technology makes it possible to produce quite elaborate animatronics at a reasonably low consumer price level.
Things might seem tough at the moment, but the recession won't last indefinitely and I expect that toys with a similar level of sophistication to Pleo will return and become much more mainstream than they are now.
Things might seem tough at the moment, but the recession won't last indefinitely and I expect that toys with a similar level of sophistication to Pleo will return and become much more mainstream than they are now.
Tuesday, April 21, 2009
Know your location
A lecture by Dieter Fox on the topic of probabilistic methods for mobile robot navigation. This is mostly concerned with finding out the robot's location at any point in time, given an existing map of the environment. Map creation and simultaneous localization and mapping are barely touched upon, but the global localization methods described - usually referred to as monte carlo localization - are still relevant and hard to beat in terms of robustness, even though this lecture is from nearly a decade ago.
Labels:
localizaion,
mobile robot,
navigation
Sunday, April 19, 2009
Penguins in the sky
Just in time for Jaunty release parties, here's a novel take on the automated blimp concept. You can find more info here.

The bionic manipulator looks interesting as a very lightweight arm design, and might be of interest for use on mobile robots.

The bionic manipulator looks interesting as a very lightweight arm design, and might be of interest for use on mobile robots.
Saturday, April 18, 2009
Sun storm
The STEREO spacecraft observe a gigantic cloud of ionized helium being ejected into space by the sun.
ESR: We don't need no GPL
Eric Raymond suggests that software licences, such as the GPL, are no longer needed.
He makes a good point, which has been borne out by experience over the last decade. Basically, companies which try to take open source software and make it closed tend to eventually lose out, because they can't keep up with the frenetic pace of bazaar style development.
This line of argument works, and there's evidence to show that it does. However, it only works for popular open source projects which have many contributors and are fairly dynamic. There are also many other open source projects which are not so popular and have only a small number of developers, or even just a single developer. For smaller projects it would be possible for a big company with many developers to take those projects, make them closed, and actually out-compete the little guys or hobbyists using their greater resources.
As I see it GPL and similar licenses help to protect these smaller scale projects from aggressive corporate takeovers. On the other hand though projects without many developers tend to be not very popular and so are less attractive to business in terms of numbers of potential customers/users, but for small projects destined to eventually become popular without a GPL-like license there would be a strong first mover advantage for corporations to grab and take them over.
He makes a good point, which has been borne out by experience over the last decade. Basically, companies which try to take open source software and make it closed tend to eventually lose out, because they can't keep up with the frenetic pace of bazaar style development.
This line of argument works, and there's evidence to show that it does. However, it only works for popular open source projects which have many contributors and are fairly dynamic. There are also many other open source projects which are not so popular and have only a small number of developers, or even just a single developer. For smaller projects it would be possible for a big company with many developers to take those projects, make them closed, and actually out-compete the little guys or hobbyists using their greater resources.
As I see it GPL and similar licenses help to protect these smaller scale projects from aggressive corporate takeovers. On the other hand though projects without many developers tend to be not very popular and so are less attractive to business in terms of numbers of potential customers/users, but for small projects destined to eventually become popular without a GPL-like license there would be a strong first mover advantage for corporations to grab and take them over.
Ubuntu 9.04 installation
Graphics driver
As usual screen resolution is always something horrible by default, 800x600 or 640x480. This is the biggest weakness of Ubuntu in my opinion, and presents the new/naive user with an immediate and serious usability problem to be overcome. Selecting all the Nvidia and proprietary drivers packages from Add/Remove allowed me to select a suitable driver which then gives a decent resolution and 3D acceleration.
Cosmetic changes
There are a couple of cosmetic differences from the previous version. The volume control is now a horizontal rather than a vertical slider. There's the much talked of changes to system notifications, which looks a little prettier, and also the default login screen is black rather than brown with a different style of logo. Under System/Preferences Screen Resolution has been renamed to Display.
MonoDevelop with integrated debugger
One big improvement from my point of view is that MonoDevelop version 2.0 is now supported, which provides an integrated debugger. However, just selecting MonoDevelop from Add/Remove is not enough to get all the functionality. Go to System/Administration/Synaptic and search for all packages related to Nunit and monodevelop-debugger, select them, then apply. On running the IDE a new toolbar button appears called "Debug", which can be used to interactively debug after setting some breakpoints.
Interactive debugging is a bonus for me because a proportion of the software I'm writing involves interaction with physical systems. It is possible to debug in the old-fashioned way by writing stuff to the console or to file, but being able to set breakpoints and watches makes life much simpler. Watching variables isn't quite as intuitive as it could be, which would be to right click and select watch from the menu, but it's easy enough to copy and paste the variable name into the list.
I'd say that MonoDevelop 2.0 is the first really usable version of this IDE. Previous versions have either been flaky or lacking in functionality compared to similar programs such as Eclipse.
Overall opinion
Another excellent release. I still occasionally use Microsoft Windows at work or on friends computers, and in my opinion contemporary Linux distros are so much better pieces of software engineering. Not having to worry about malware/viruses and all that crap or needing to install cumbersome "security" ransomware is a big advantage, and everything runs so much more quickly. The installation process for Ubuntu is slick and looks professional, let down only by it being unclear how to increase the screen resolution. For someone like me with a lot of computing experience this is a trivial issue, but for a first time user it could be a show stopper.
As usual screen resolution is always something horrible by default, 800x600 or 640x480. This is the biggest weakness of Ubuntu in my opinion, and presents the new/naive user with an immediate and serious usability problem to be overcome. Selecting all the Nvidia and proprietary drivers packages from Add/Remove allowed me to select a suitable driver which then gives a decent resolution and 3D acceleration.
Cosmetic changes
There are a couple of cosmetic differences from the previous version. The volume control is now a horizontal rather than a vertical slider. There's the much talked of changes to system notifications, which looks a little prettier, and also the default login screen is black rather than brown with a different style of logo. Under System/Preferences Screen Resolution has been renamed to Display.
MonoDevelop with integrated debugger
One big improvement from my point of view is that MonoDevelop version 2.0 is now supported, which provides an integrated debugger. However, just selecting MonoDevelop from Add/Remove is not enough to get all the functionality. Go to System/Administration/Synaptic and search for all packages related to Nunit and monodevelop-debugger, select them, then apply. On running the IDE a new toolbar button appears called "Debug", which can be used to interactively debug after setting some breakpoints.
Interactive debugging is a bonus for me because a proportion of the software I'm writing involves interaction with physical systems. It is possible to debug in the old-fashioned way by writing stuff to the console or to file, but being able to set breakpoints and watches makes life much simpler. Watching variables isn't quite as intuitive as it could be, which would be to right click and select watch from the menu, but it's easy enough to copy and paste the variable name into the list.
I'd say that MonoDevelop 2.0 is the first really usable version of this IDE. Previous versions have either been flaky or lacking in functionality compared to similar programs such as Eclipse.
Overall opinion
Another excellent release. I still occasionally use Microsoft Windows at work or on friends computers, and in my opinion contemporary Linux distros are so much better pieces of software engineering. Not having to worry about malware/viruses and all that crap or needing to install cumbersome "security" ransomware is a big advantage, and everything runs so much more quickly. The installation process for Ubuntu is slick and looks professional, let down only by it being unclear how to increase the screen resolution. For someone like me with a lot of computing experience this is a trivial issue, but for a first time user it could be a show stopper.
Wednesday, April 15, 2009
Unethical uses of robotics
Are iRobot employees part of the solution, or part of the problem? I once looked up to people like Colin Angle, but perhaps I was just completely mistaken.
Is stuff like this improving the future, or making it more volatile and dangerous?

It was only a matter of time before those iRobots transitioned from fairly uncontroversial applications such as bomb disposal and rescue into being offensive weapons. They're still just telerobots, but once the mechanical engineering problems are solved it's the logical next step to make it a semi autonomous and then fully autonomous offensive weapon.
Of course there's a lot of hype about these being used to remove troops from battlefields, reduce casualties, and so on. But anyone with an ounce of sense knows that soon non-state actors are also going to be using these, so you better get acclimatized to the notion of a hundred, or a thousand, of these driving into a city centre near you and causing chaos. Modern police forces and civilian emergency agencies are just totally unprepared for this kind of danger. You can't arrest an armed telerobot. You can only disable or destroy it.
Is stuff like this improving the future, or making it more volatile and dangerous?

It was only a matter of time before those iRobots transitioned from fairly uncontroversial applications such as bomb disposal and rescue into being offensive weapons. They're still just telerobots, but once the mechanical engineering problems are solved it's the logical next step to make it a semi autonomous and then fully autonomous offensive weapon.
Of course there's a lot of hype about these being used to remove troops from battlefields, reduce casualties, and so on. But anyone with an ounce of sense knows that soon non-state actors are also going to be using these, so you better get acclimatized to the notion of a hundred, or a thousand, of these driving into a city centre near you and causing chaos. Modern police forces and civilian emergency agencies are just totally unprepared for this kind of danger. You can't arrest an armed telerobot. You can only disable or destroy it.
Labels:
military robots,
telerobotics
Sunday, April 12, 2009
Email schmeemail
Scurrilous emails are a common occurrence, but some politicians still seem to be labouring under the illusion that emails can be "private" (what a quaint notion). No matter how much you might try to delude yourself with "security" systems and wotnot, unless you use strong encryption on all of your email communications any messages which you send by email might as well be announced in public using a megaphone.
Especially in the UK, but also elsewhere in the world, there are now numerous systems at numerous levels within the network intercepting, filtering and data mining email traffic, and anything which might be politically embarassing must have a high monetary value attached to it because you can then sell it to the newspapers.
The Guardian goes into a bit more detail about the murky world of politicians rather hamfistedly trying to influence bloggers. Bloggers can of course be bought off like anyone else, but in the long run it's going to be much harder for the government to control the opinions which are aired within the blogosphere in the same way that they controlled some the popular press and television - the "old media" - in the past.
Especially in the UK, but also elsewhere in the world, there are now numerous systems at numerous levels within the network intercepting, filtering and data mining email traffic, and anything which might be politically embarassing must have a high monetary value attached to it because you can then sell it to the newspapers.
The Guardian goes into a bit more detail about the murky world of politicians rather hamfistedly trying to influence bloggers. Bloggers can of course be bought off like anyone else, but in the long run it's going to be much harder for the government to control the opinions which are aired within the blogosphere in the same way that they controlled some the popular press and television - the "old media" - in the past.
Saturday, April 11, 2009
CCTV cars
Continuing the ongoing British obsession with spying on people's activities is a new generation of "CCTV cars". Cameras attached to police vehicles is nothing new, but this system extends the capabilities beyond fixed mountings using a telescopically elevated pan and tilt mechanism.
Anyway, the main point of this seems to be detecting minor traffic violations, since these result in fines which supplement the income of police agencies and local authorities. Such revenue streams may be regarded as especially important during the recession as money from central government budgets is reduced. It's also possible to use this to look through the upstairs windows of houses, over walls, hedges or fences into private gardens, and so on.
There is no escape for minor offenders from the telescopically long arm of the law! Crisp eating channel hoppers and topless garden sunbathers, you have been warned.
"Anyone seen driving while distracted - eating at the wheel, playing with the radio or applying make-up for instance - is filmed by the cameras"Changing radio channels whilst driving is now a crime?? Don't touch that dial!
Anyway, the main point of this seems to be detecting minor traffic violations, since these result in fines which supplement the income of police agencies and local authorities. Such revenue streams may be regarded as especially important during the recession as money from central government budgets is reduced. It's also possible to use this to look through the upstairs windows of houses, over walls, hedges or fences into private gardens, and so on.
There is no escape for minor offenders from the telescopically long arm of the law! Crisp eating channel hoppers and topless garden sunbathers, you have been warned.
OpenCog screencasts
Some screencasts showing aspects of the OpenCog system. These are clearer than other screencasts which I've seen, and the audio quality is much better too.
http://fruitionnz.com/dist/OpenCog-ImportanceUpdatingAgent.ogv (~80M)
http://fruitionnz.com/dist/OpenCog-HebbianUpdatingAgent.ogv (~55M)
http://fruitionnz.com/dist/OpenCog-ImportanceDiffusionAgent.ogv (~75M)
http://fruitionnz.com/dist/OpenCog-ForgettingAgent.ogv (~15M)
http://fruitionnz.com/dist/
http://fruitionnz.com/dist/
http://fruitionnz.com/dist/
http://fruitionnz.com/dist/
Inflation, not deflation
There has been a lot of talk in the media about the "great danger of deflation". But actually deflation at the moment isn't really a problem - quite the opposite in fact.
These graphs are for the US, but I bet the graphs here look similar. Just from keeping track of my own weekly shopping, prices have risen noticeably over the last couple of years, especially on items like bread and fruit. Inflating prices are an inevitable consequence of all the money printing, aka "quantitative easing", going on and being actively promoted by the government. Pumping valueless currency into the economy just dilutes the value of your money, so that every pound or dollar you spend is worth less.
The biggest inflation has been in energy - heating gas, fuel oil and electricity - but this can be partly explained by the oil price spike which took place in 2008. Over the last year by gas bill rose by 55.4%, and the electricity bill rose by 178.2%. In theory energy prices should come down eventually, but at present they remain high.
These graphs are for the US, but I bet the graphs here look similar. Just from keeping track of my own weekly shopping, prices have risen noticeably over the last couple of years, especially on items like bread and fruit. Inflating prices are an inevitable consequence of all the money printing, aka "quantitative easing", going on and being actively promoted by the government. Pumping valueless currency into the economy just dilutes the value of your money, so that every pound or dollar you spend is worth less.
The biggest inflation has been in energy - heating gas, fuel oil and electricity - but this can be partly explained by the oil price spike which took place in 2008. Over the last year by gas bill rose by 55.4%, and the electricity bill rose by 178.2%. In theory energy prices should come down eventually, but at present they remain high.
Friday, April 10, 2009
Be open minded and exercise skepticism
I often wish that people would treat the claims made by politicians and in newspapers with more skepticism. For instance, a couple of days ago on the radio I heard that the government's terror threat status was something like "über critical", meaning that they want you to believe that a major terrorist incident is imminent within hours or days. But is this really true? If so, where's the evidence? If they don't have any evidence, or they claim that they do have evidence but won't tell you what it is even in very non-specific terms, this probably means that the claim itself is without merit. If you have no way of evaluating whether some assertion is true or false, it's usually safer to assume that it's false until more evidence comes to light.
Storing occupancy grids
One of the problems with 3D occupancy grids, is that they can occupy a lot of storage space in terms of memory or disk storage. Imagine a space the size of an average house turned into small 1cm cubes, and that's quite a lot of cubes to keep track of.
Much of the space inside homes is actually empty, or rather filled with air, but from the robot's point of view knowing about probably empty space is just as important (maybe even more important!) than knowing about what is occupied, and thereby a potential obstacle. Some savings can be made by not storing information about terra incognita - areas of the map which have so far not been explored, but assuming that we want the robot to have a good understanding of an entire house this still leaves us with quite a heap of data.
At this point the unimaginative can simply appeal to Gordon Moore and his famous "law". The capacity of storage devices, such as hard disk drives, is always increasing and it does look as if even the smallest storage devices around today would be able to handle the number of cubes that we would like to deal with. Even though this is the case loading from and saving to the storage device is still going to be relatively slow, and the robot needs to be able to access the data more or less in real time if it's going to be useful. We could also be lazy and just load the whole lot into a large amount of RAM, but ideally it would be good if low cost devices could be used, such as netbooks, which only have modest memory and local storage capacity. This would help robotics to continue becoming more economical and therefore marketable.
So what to do? Since the occupancy data in this case is being produced from stereo vision a way to get better storage economy might be to only store a random sample of the stereo disparities observed from a dense disparity image. If we know the location and pose from which the observation was originally made, based upon the results of SLAM, then a local 3D occupancy grid can be regenerated dynamically from a fairly small amount of data as the robot moves around the house. This means that storage access times are going to be much shorter, and potentially a lot of stereo disparity data could be buffered in memory.
Some back an envelope calculations go as follows:
If we randomly sample 300 stereo disparities from a dense disparity image, and represent the image coordinates and disparity as floating point values (sub-pixel accuracy), this translates into
300 stereo features x 3 values (x,y,disparity) x 4 bytes per value
= 3600 bytes per observation, or 3.5K
If we also want to store colour information, so that coloured 3D occupancy grids can be produced this increases to 4500 bytes or 4.4K. There is also the robot's pose information to store, but this is only a small number of bytes, so doesn't make a big overall difference. This seems quite tractable. Potentially the robot could make several thousand observations as it maps the house, and this only translates into a few tens of megabytes which is well within the limitations of what a netbook could handle. Even if the number of observations rises into the tens of thousands this still looks feasible.
Additionally, if the disparities are saved to disk in a compressed format, using a fast compression method, we can improve the storage efficiency even further. This does add some computational overhead, but when you balance this against the time spent reading from and writing to a physical storage media in uncompressed form this can be a bargain worth striking.
Much of the space inside homes is actually empty, or rather filled with air, but from the robot's point of view knowing about probably empty space is just as important (maybe even more important!) than knowing about what is occupied, and thereby a potential obstacle. Some savings can be made by not storing information about terra incognita - areas of the map which have so far not been explored, but assuming that we want the robot to have a good understanding of an entire house this still leaves us with quite a heap of data.
At this point the unimaginative can simply appeal to Gordon Moore and his famous "law". The capacity of storage devices, such as hard disk drives, is always increasing and it does look as if even the smallest storage devices around today would be able to handle the number of cubes that we would like to deal with. Even though this is the case loading from and saving to the storage device is still going to be relatively slow, and the robot needs to be able to access the data more or less in real time if it's going to be useful. We could also be lazy and just load the whole lot into a large amount of RAM, but ideally it would be good if low cost devices could be used, such as netbooks, which only have modest memory and local storage capacity. This would help robotics to continue becoming more economical and therefore marketable.
So what to do? Since the occupancy data in this case is being produced from stereo vision a way to get better storage economy might be to only store a random sample of the stereo disparities observed from a dense disparity image. If we know the location and pose from which the observation was originally made, based upon the results of SLAM, then a local 3D occupancy grid can be regenerated dynamically from a fairly small amount of data as the robot moves around the house. This means that storage access times are going to be much shorter, and potentially a lot of stereo disparity data could be buffered in memory.
Some back an envelope calculations go as follows:
If we randomly sample 300 stereo disparities from a dense disparity image, and represent the image coordinates and disparity as floating point values (sub-pixel accuracy), this translates into
300 stereo features x 3 values (x,y,disparity) x 4 bytes per value
= 3600 bytes per observation, or 3.5K
If we also want to store colour information, so that coloured 3D occupancy grids can be produced this increases to 4500 bytes or 4.4K. There is also the robot's pose information to store, but this is only a small number of bytes, so doesn't make a big overall difference. This seems quite tractable. Potentially the robot could make several thousand observations as it maps the house, and this only translates into a few tens of megabytes which is well within the limitations of what a netbook could handle. Even if the number of observations rises into the tens of thousands this still looks feasible.
Additionally, if the disparities are saved to disk in a compressed format, using a fast compression method, we can improve the storage efficiency even further. This does add some computational overhead, but when you balance this against the time spent reading from and writing to a physical storage media in uncompressed form this can be a bargain worth striking.
Tuesday, April 07, 2009
Automotive X-prize vehicles
Some of the vehicles entered into the automotive X-prize.
I think it would be a good idea for contestants to try to remain with a chassis design which looks somewhat conventional. I would happily drive some of the far out and futuristic-looking vehicles, because I don't care what the car itself looks like, but my attitude towards these things seems not to be the norm. Even if a superb car was produced with excellent energy efficiency it might still fail in the market because to many people "it doesn't look like a real car".
Also I think the seating position should remain conventional. The more laid back, sports car type seating wouldn't suit many drivers. Think of how disabled people, pensioners or mums taking children to school might use the vehicle, and that's probably a good design guide. If the car design is too adventurous then at best it will just remain a toy for a wealthy minority, only a short step removed from a "concept car".
I think it would be a good idea for contestants to try to remain with a chassis design which looks somewhat conventional. I would happily drive some of the far out and futuristic-looking vehicles, because I don't care what the car itself looks like, but my attitude towards these things seems not to be the norm. Even if a superb car was produced with excellent energy efficiency it might still fail in the market because to many people "it doesn't look like a real car".
Also I think the seating position should remain conventional. The more laid back, sports car type seating wouldn't suit many drivers. Think of how disabled people, pensioners or mums taking children to school might use the vehicle, and that's probably a good design guide. If the car design is too adventurous then at best it will just remain a toy for a wealthy minority, only a short step removed from a "concept car".
Monday, April 06, 2009
914s spoon-fed crackers and spinich by a bomb disposal robot
This was so odd that it just had to be included.
Sunday, April 05, 2009
Maybe good news on nuclear disarmament
Amidst the popular gloom this does look like a piece of good news, because for well over a decade any significant reductions in nuclear weapons capability has not even been on the political agenda.
By far the biggest threat from nuclear weapons still comes from the US and Russia, who each have thousands of aging nuclear missiles and bombs left over from the cold war of the 20th century. One of the biggest concerns when large numbers of weapons are stockpiled like this arises from the potential of an accidental explosion to trigger a global war, in accordance with the automaticity of the mutually assured destruction doctrine. An accident could easily be interpreted as a deliberate attack by a hostile enemy, especially considering the numbers of casualties and the environmental damage which might be incurred.
Many of these weapons and their accompanying control systems will be decades old, and the engineers who built them may have moved on to other jobs or retired. When times get tough there is also the temptation to cut corners when it comes to checks, maintenance, and security. Of course the procedures in place are supposed to be stringent, but we all know that officials can be corrupted and that filling in forms with bogus information in order to fulfill the demands of quotas or performance targets seems to be quite common amongst government bureaucracies, especially on a tight budget.
According to the news article both the US and Russia will be aiming to reduce their arsenals to merely 1500 nukes each. If you conservatively estimate that each of these is more powerful than the bombs dropped on Japan at the end of WW2, and that if used 50% of them would fail to detonate or reach their target each of the superpower nations could still destroy 750 cities on the opposing side. Sum the populations of the 750 largest cities in the US and in Russia, and that provides a good indication of how many lives are potentially at stake. In other words, even if the claimed reductions are achieved there will still be more than enough weapons around to bring an abrupt end to civilization and progress as we currently know it.
By far the biggest threat from nuclear weapons still comes from the US and Russia, who each have thousands of aging nuclear missiles and bombs left over from the cold war of the 20th century. One of the biggest concerns when large numbers of weapons are stockpiled like this arises from the potential of an accidental explosion to trigger a global war, in accordance with the automaticity of the mutually assured destruction doctrine. An accident could easily be interpreted as a deliberate attack by a hostile enemy, especially considering the numbers of casualties and the environmental damage which might be incurred.
Many of these weapons and their accompanying control systems will be decades old, and the engineers who built them may have moved on to other jobs or retired. When times get tough there is also the temptation to cut corners when it comes to checks, maintenance, and security. Of course the procedures in place are supposed to be stringent, but we all know that officials can be corrupted and that filling in forms with bogus information in order to fulfill the demands of quotas or performance targets seems to be quite common amongst government bureaucracies, especially on a tight budget.
According to the news article both the US and Russia will be aiming to reduce their arsenals to merely 1500 nukes each. If you conservatively estimate that each of these is more powerful than the bombs dropped on Japan at the end of WW2, and that if used 50% of them would fail to detonate or reach their target each of the superpower nations could still destroy 750 cities on the opposing side. Sum the populations of the 750 largest cities in the US and in Russia, and that provides a good indication of how many lives are potentially at stake. In other words, even if the claimed reductions are achieved there will still be more than enough weapons around to bring an abrupt end to civilization and progress as we currently know it.
Saturday, April 04, 2009
Thursday, April 02, 2009
Can birds count?
A brief and it has to be admitted unconvincing article suggests that newly hatched chicks can count up to three. I also remember reading in the past that parrots may be able to count small numbers of objects. These sorts of experiments with animals need to be done carefully, but under the right conditions it should be possible to rule out the clever Hans effect, where the animal has just learned to respond to a particular tone of voice or facial expression.
Might there be any adaptive value for birds to be able to count? Well, if you're guarding a nest it might be useful to be able to keep track of how many eggs are there. If one goes missing then adopting a more vigilant "guard mode" would be adaptive, as too would be noticing if a new egg has unexpectedly appeared (for example laid by a cuckoo).
Might there be any adaptive value for birds to be able to count? Well, if you're guarding a nest it might be useful to be able to keep track of how many eggs are there. If one goes missing then adopting a more vigilant "guard mode" would be adaptive, as too would be noticing if a new egg has unexpectedly appeared (for example laid by a cuckoo).
Subscribe to:
Posts (Atom)