Robotics : state of the art technologies and hobby

[William Gaatjes]

I thought it would be nice to make a robotics thread.
To post here the latest achievements.
Please post findings if you have them...

We all know the amazing robots if Boston dynamics...
Such as the alpha dog.
http://www.youtube.com/watch?v=SSbZrQp-HOk

Today i read about a robotic flying orb from a research team for the defense department in Japan...

http://www.youtube.com/watch?v=pF0uLnMoQZA

 

[William Gaatjes]

Here is a nice website explaining all the basics about sensors and more...
http://www.societyofrobots.com/sensors.shtml

 

[Modelworks]

The snake robot is one of my favorites
http://www.youtube.com/watch?v=8VLjDjXzTiU

 

[William Gaatjes]

The snake robot is one of my favorites
http://www.youtube.com/watch?v=8VLjDjXzTiU

With more segments it would be even a better climber then it is already.

Imagine what we could build when we finally have a proper power source...
When we finally have a fuel cell or battery technology that can rival the energy storage capacity of carbohydrates for the same amount of size en weight, humanity will leap forward. After that the advances of even better power generation and distributions will be the driving force. More powerful actuators...

Sigh... I cannot wait...

 

[Matthiasa]

Definitely beats the project I have going for a class of mine.
Dumb sensors, motor, inverters(motor control), batteries, etc... If only had the budget (and time) for good stuff.

 

[Jeff7]

RHex - nifty little thing.

Alternate mobility mode.

 

[The Boston Dangler]

RHex - nifty little thing.

Alternate mobility mode.

BigDog Weaponized

 

[William Gaatjes]

BigDog Weaponized

Yeehaw ! :biggrin:

A movie about a robot that can changes it's wheels into legs.
It drives around, or it runs around like a beetle to get past terrain where wheels are of no use.This robot has been created by Bio-Inspired Robotic Laboratory (BioRoLa) of National Taiwan University.

http://www.youtube.com/watch?&v=a9N6Zb_K2Z0

 

[exdeath]

Just reminds me of the two reasons why our technology sucks and why being stuck on this rock is depressing.

1) data storage - we still use magnetic read/write heads on spinning disks and tape to record data at mere megabytes per second speeds and this hasn't changed in over 50 years. Flash memory isn't much better. A modern CPU can process 100s of GB per second but we store that data using technology that was already crude in the 1950s at 10 MB/sec random access. To be able to do things like computer vision and AI or implement any semblance of intelligence or consciousness requires recalling and comparing gigabytes and terabytes of memories and experiences in fractions of a second. Our processing power is nearly there, but our ability to store and retrieve non volatile data is an absolute joke.

2) power sources - goes without saying. Chemical cell power sources, or chemical based energy in general, have long been inadequate and obsolete.


We can build the fastest CPU/MCU in the world, and in the end we are still relying on a guy with a hammer and chisel to store and retrieve the working data serially one bit at a time, connect it to other systems with 2 cups and a string, and power it all with fire, steam, and acids. Too primitive for a species that prides itself on it's "technological achievement".

We need some serious breakthroughs in physics if life is going to be interesting at all in the next 100 years; we've pretty much hit a brick wall with current knowledge. Well just keep making faster Exaflop trillion core CPUs so we can insert more wait states to read data from the storage device at a whopping 10 mb/sec and killing batteries in 15 minutes doing it.

 

[exdeath]

Imagine what we could build when we finally have a proper power source...
When we finally have a fuel cell or battery technology that can rival the energy storage capacity of carbohydrates for the same amount of size en weight, humanity will leap forward.

Likewise for data storage.

Eliminating the memory gap with a true 1:1 non volatile universal memory (eg: storing and retrieving data from the primary data store as fast as a current CPU can execute mov instructions) would kick start a revolution in computing power greater than that of the transistor and microprocessor combined. Processors are just so damned fast that everything today is IO bottlenext by 5 decade old storage technologies.

I doubt the human brain is constricted to 1 MB/sec 4k random access when processing real time 3D image data from the retinas, recounting memories, images, conversations, comparing, storing, etc, in the blink of an eye.

 

[Modelworks]

Not completely without human control but lots of robotic engineering at work in probably one of the most entertaining robots.
http://www.youtube.com/watch?v=qPsqsPgC9v4&feature=related

 

[irishScott]

http://www.youtube.com/watch?v=GQ3AcPEPbH0

Never gets old.

 

[William Gaatjes]

Likewise for data storage.

Eliminating the memory gap with a true 1:1 non volatile universal memory (eg: storing and retrieving data from the primary data store as fast as a current CPU can execute mov instructions) would kick start a revolution in computing power greater than that of the transistor and microprocessor combined. Processors are just so damned fast that everything today is IO bottlenext by 5 decade old storage technologies.

I doubt the human brain is constricted to 1 MB/sec 4k random access when processing real time 3D image data from the retinas, recounting memories, images, conversations, comparing, storing, etc, in the blink of an eye.

I agree with the limitations between memory and processing.
What we need is processing directly on memory(This is what the brain does, it incorporates the logic and the memory, as if it is a sort of reconfigurable fpga). One commercial method is used in the current xbox 360 with the 10mb edram. For example, the TI arm chip on the beagleboard xm, uses pop technology to directly connect 512MB of dram memory on the mcu die. Very short traces, And can be a very wide data path if needed.

This is the key issue i think :
Only read and only write processes can be cached easily.
But read modify write processes (modifying data) are very bandwidth intensive(at least 2x bandwidth) and should be designed to be as close to memory as possible or should be modified that reading is no longer needed when modifying data. That will ask for logic on board memory and for example just as the Atmel peripheral logic of the SAM7S series separate buses and seperate ports to only write 0 or to only write 1. Which are or-ed together to modify the data. Moving logic to the memory is the next logical step. But i am sure that patent systems such as the current system is killing innovation. Similar as the Rambus dram debacle in the past.

I think that in the near future, multi layer memory bus systems will become more apparent connected to memory. But this will also ask different programming models and a more structured way of storing data. Now data is just literally dumped somewhere in ram. And that will need to change.

Another example is dual port memory. But the problem is that this is difficult to implement with dram technology. Simply because reading or writing dram memory asks for a refresh of the capacitor. Thus needing large buffers but always with a delay if a memory bank cannot be accessed.

But with memristor technology becoming more mature, things are going to change soon, i think we can safely assume.

 

[William Gaatjes]

Current state of memristors memory technology :

http://www.dailytech.com/HP+to+Depl...Replacement+Within+18+Months/article22963.htm

October 10, 2011 9:17 AM

HP to Deploy Memristor Powered SSD Replacement Within 18 Months

HP's thin film memristor process, which creates components at a 5 nm scale, can deposit 500 billion memristors within a single chip layer. What this means is that layers of memristor non-volatile storage (akin to NAND flash) and memristor RAM can be placed directly atop the CPU and GPU cores in system-on-a-chip designs. Using faster direct interconnects and scrapping the system bus for anything other than I/O, these future memristor enabled SoCs could deliver drastic battery life and processing power leaps over the hardware currently found in smartphones, tablets, and laptops.

Comments Mr. Williams, "We put the non-volatile memory right on top of the processor chip, and, because you’re not shipping data off-chip, that means we get the equivalent of 20 years of Moore’s Law performance improvement. We’re running hundreds of wafers through the fab. We’re way ahead of where we thought we would be at this moment in time."

HP currently owns a key patent [U.S. Patent 7443711] on its memristor implementation and the process to build it. And it has more patents pending. Mr. Williams says HP's plan is not to return to the components business where it was once a strong player, but rather to license the technology to interested components manufacturers.

He comments, "We're the world’s largest purchaser of DRAM and the second largest buyer of flash and we’re trying to disrupt and re-arrange our supply chain. The plan is to license this technology to anyone who wants it, and we'll teach them how to make it. But you'll have to stand in line, we have a bunch of people queued for it. We're doing this because, frankly, we didn't see a hell of a lot of innovation happening out there."

http://www.hpl.hp.com/news/2008/apr-jun/memristor_faq.html

Memristance is a property of an electronic component. If charge flows in one direction through a circuit, the resistance of that component of the circuit will increase, and if charge flows in the opposite direction in the circuit, the resistance will decrease. If the flow of charge is stopped by turning off the applied voltage, the component will 'remember' the last resistance that it had, and when the flow of charge starts again the resistance of the circuit will be what it was when it was last active.

It turns out that memristance is becoming stronger as the feature sizes in circuits are getting smaller. At some point as we scale into the realm of nanoelectronics, it will be necessary to explicitly take account of memristance in our circuit models in order to simulate and design electronic circuits properly.

 

[exdeath]

Would be nice. Human brain is like having 1000+ TB of battery backed SRAM that lasts 60+ years. Zero wait state.

 

[sm625]

exdeath, if you got the money we got the technology. You can take an XMC module with a xilinx virtex 6 FPGA and write your own custom code and get high memory bandwidth and low latency. You can build a system using dozens or even hundreds of these modules. The technology is there, the software is lagging. We have more tech than we know what to do with. And the problem is only getting worse because too many people are sitting around watching fail videos instead of being creative.

Look at this thing:

http://www.4dsp.com/VID675.php

That kind of technology was unheard of even 10 years ago. The amount of image processing that can be done with that is far beyond what all but a very few people can even conceive of using.

Be careful what you wish for. While most of us may not even know how far along this stuff is, there are some who do know, and they arent necessarily working for the betterment of humanity.

 

[Jeff7]

Would be nice. Human brain is like having 1000+ TB of battery backed SRAM that lasts 60+ years. Zero wait state.

- And it's also saddled with compression that ranges from nearly-lossless, all the way down to a JPEG that's been reopened and then resaved 50x at the 0 quality setting. The bonus is that you have little control over what type of compression is going to be used. Writing things to permanent memory also requires anywhere from one to a dozen refreshes.
- There's also no ECC support.
- The compression used on stored information does tend to become increasingly lossy as time progresses.
- Random access of information is also spotty, due to poor indexing functionality.
- It's very susceptible to radiation damage.
- Irreversible damage occurs within minutes of a power interruption. (Not just data loss - permanent damage. )
- Low G-force tolerance.
- No upgradability. (With current technology.)
- No user manual.


+ High capacity for a few specific types of parallel computing.


If it was a computer, I'd want a refund.

 

[William Gaatjes]

- And it's also saddled with compression that ranges from nearly-lossless, all the way down to a JPEG that's been reopened and then resaved 50x at the 0 quality setting. The bonus is that you have little control over what type of compression is going to be used. Writing things to permanent memory also requires anywhere from one to a dozen refreshes.
- There's also no ECC support.
- The compression used on stored information does tend to become increasingly lossy as time progresses.
- Random access of information is also spotty, due to poor indexing functionality.
- It's very susceptible to radiation damage.
- Irreversible damage occurs within minutes of a power interruption. (Not just data loss - permanent damage. )
- Low G-force tolerance.
- No upgradability. (With current technology.)
- No user manual.


+ High capacity for a few specific types of parallel computing.


If it was a computer, I'd want a refund.

I know you are making a joke, but most of your arguments are a really big fail.

My opinion :
ECC is not needed when the redundancy is stored as part of the information.
The indexing is also part of the stored information.
While at the same time removing any form of data that is useless. Culling of noise. The reason why the brain needs to do data noise culling up front is because the noise has a very negative effect on indexing. Just ask people to use imagination when looking at a random pattern of dots. Index window of the internal mind working over time during creativity and the results of that index window are determined by :

• The neurotransmitters present.
• Signals from senses.
• The internal virtual senses. The groups of neurons that act as if there are senses, but in reality create virtual sensor data. This gives as the ability to think and to simulate environmental situations that did not happen. This is looking ahead, predicting based on previous stored information and changing the variables a bit to simulate a different outcome.

The data compression of the brain is inversely proportional to the repetition rate of the data. The more often a piece of information is presented, the more accurate it is stored.
The side effect is that the more accurate the information is stored, the higher the index value becomes, meaning this information attains a higher priority.

Because of the large form of redundancy, the brain is self healing. And with presenting the same data as what was previously stored but damaged because of whatever reason(for example radiation), the data can be corrected (Examples : Learning to walk again, learning to speak again, learning to do a certain task again). Raid 5 type storage for example could be very similar as how data is stored in the brain. Although not completely, storing bits or storing information is something completely different.

The brain just works at a radical different level of information storage technology when compared to the modern hardware/ software combo.
But it will become in the near future very easy to implement typical brain functions.

p.s. :

Any one who is interested, read up how gpu cards do polygon culling, and what the gpu & edram can do in the xbox360. It would be a nice project for advanced AI with fpga logic or multi core processing. The brain is not a classical database system. And that is why most ai implementations do not or will work. The magic can be found in how the gpu processes data to reduce the information only to what is really needed for the view port.

 

[William Gaatjes]

I received this link from a friend.

It is about a small robot driving a bicycle.
http://www.youtube.com/watch?v=SqBw7XapJKk&


Amazing ! And funny to watch too. Kind of cute...:wub:

 

[CycloWizard]

exdeath, if you got the money we got the technology. You can take an XMC module with a xilinx virtex 6 FPGA and write your own custom code and get high memory bandwidth and low latency. You can build a system using dozens or even hundreds of these modules. The technology is there, the software is lagging. We have more tech than we know what to do with. And the problem is only getting worse because too many people are sitting around watching fail videos instead of being creative.

Look at this thing:

http://www.4dsp.com/VID675.php

That kind of technology was unheard of even 10 years ago. The amount of image processing that can be done with that is far beyond what all but a very few people can even conceive of using.

Be careful what you wish for. While most of us may not even know how far along this stuff is, there are some who do know, and they arent necessarily working for the betterment of humanity.

That's still a severe bandwidth limitation if we're talking about high-framerate applications. For example, if I am using an 8-bit camera with 640x480 resolution at 10kfps, the bandwidth is 24.6 Gbps - two orders of magnitude higher than that system can tolerate. There are plenty of applications for high framerate machine vision but the technology isn't there yet to enable them. This is something I've been working on a bit recently and the solution we proposed is to construct an ASIC with the image processing algorithms in hardware. It's an expensive solution, but I couldn't find any other way to get the job done using existing technology. We could stack a huge number of FPGA systems together (close to 100 for the example above), but there are lots of overhead issues involved.

 

[bwanaaa]

http://www.youtube.com/watch?v=_sUeGC-8dyk&feature=player_embedded

this is where robots excel. they cooperate.

unlike humans.

 

[William Gaatjes]

this is where robots excel. they cooperate.

unlike humans.

:biggrin::biggrin:

That is amazingly funny to watch.
Especially watching those Quadrocopters playing the piano.

It reminds me of the little baby robots from this movie : "batteries not included"

http://en.wikipedia.org/wiki/*batteries_not_included





The scene with the boxer... :'\ sigh...


Imagine that not so long from now we could have autonomous robots programmed to move and mimic pets but have still their own unique personality.

Perhaps cyclowizard will make it happen... ^_^

 

[William Gaatjes]

Another movie about the flying orb from the Japanese defense department.
It has been mentioned that the total cost is around 1500 dollar for parts.
The magic is of course in the hardware & software integration...

This reminds me of Half life 2. The drones following...



http://www.youtube.com/watch?v=2RK_rl6oXfo
The most cutest reporter.

 

[wirednuts]

oh noes the japanese got r/c planes now! were fucked!

 

[videogames101]

Same team I believe:

http://www.youtube.com/watch?v=YQIMGV5vtd4