Dec 10, 2014

Sensor Algorithms Go Open Source

posted by Bryon Moyer

So you’re bringing together some sensors in your system, and you want to fuse their data together. Congratulations; you’re in good company. You and the rest of the world want this. If this capability is simply a need-to-have, while not being the key differentiator for you, well, yay, you get to reinvent the fusion algorithms that everyone else has already reinvented. Or you can buy them with dollars that your Board would rather be spent on differentiating technology.

Or, let’s say this capability is a differentiator for you – you’ve got some new algorithmic ideas that you think will put you ahead. Well, good luck. In addition to the 15% of code change that you want to do, you’ll have to do the other 85% from scratch. Or buy source code from someone else (if they’ll let you have it) and learn their code and make the mods yourself. Or have them make the mods – and turn over your differentiating ideas to them.

Well, now there’s another option: the Accelerated Innovation Community (AIC), under the aegis of the MEMS Industry Group (MIG). I know… the name doesn’t say much about what it is, but the idea is to have an open-source community where sensor-fusion algorithms can be shared. This means that you can get ready-working algorithms easily. They may not work exactly the way you want, but they make a great starting point.

If you find ways to improve them, you can give back to the community. And if you don’t want to give back differentiating improvements, you don’t have to, since the algorithms are available under BSD licensing.

To be clear, these aren’t necessarily shrink-wrapped algorithms that can be plugged into some abstracted system with no engineering work. You will still need algorithm coders and embedded folks that know what they’re doing. The benefit of this is that they won’t have to start from scratch every time.

Now… if you’ve been paying attention, you’ll say, “But wait! What about that open sensor platform (OSP) thing from earlier this year?” Yes, your memory is fine: that happened. And these two things don’t compete: in fact, they can work together.

The OSP piece is a framework onto which algorithms can be hung. Any algorithms, including ones obtained through the AIC, can be used. Think of it as a standard API you would use to invoke the algorithms (or do other things – OSP is about more than just algorithms).

You also might wonder, given that Audience has a different (or more focused) business than Sensor Platforms did, will they still support OSP? The answer is, “Yes.” In fact, look for a new OSP revision early next year.

By the way, at the risk of yet more confusion, note that the OSP bag of goodies contains a few reference algorithms as well. These are helpful in implementing the framework, but are not the focus of OSP; the framework is.

Freescale_figure.pngJust as Sensor Platforms (now part of Audience) and ARM were founding members of the OSP, so Freescale is an inaugural member of the AIC. Coincident with their support of the AIC, they’ve also put forward a sensor fusion dev kit. It can act as its own framework, and also contains reference algorithms. 

Freescale's kit isn’t part of the AIC, but can accept algorithms from the AIC. And they've contributed algorithms to the AIC. That’s kind of the point: algorithms should be pluggable (with the appropriate software ministrations) into any environment – whether OSP or a proprietary setup. If not, then you might have a partitioning issue that would be worth addressing…

You can find out more about the AIC in the MIG announcement.

 

(Image courtesy Freescale)

Tags :    0 comments  
Dec 09, 2014

Managing Context Options

posted by Bryon Moyer

Complexity_image.jpgThere are so many stories about the ways in which ubiquitous sensors and The Cloud will transform our lives. Most of these stories come in the form of small examples that illustrate just how powerfully this can affect what we do on a day-to-day basis.

The idea, of course, is that The Cloud knows lots of things about you – what you like, where you go, what’s on your calendar, who your friends are, and on and on – and can anticipate your needs or do things for you. Ideally, you don’t even have to think about it: it just happens

I’m a bit more skeptical; as someone who has always felt a sense of “otherness,” machines tend to make decisions for me that aren’t what I would make for myself. And I know I’m not alone here. So I would need a lot of proof that the automatic rules established on my behalf by some Greater Cumulonimbal Intelligence really were getting things right before handing over control.

And if the rules aren’t working? Then you should be able to set them yourselves. The only thing more frustrating than rules that get it wrong are ones where you can’t override them. So having, for example, phone settings would be a useful backup. Some people might never need them; others might rely on them a lot.

But a presentation at ICCAD last month got me thinking. The speaker described the scenario of going to a hospital – in the context of a phone that reports where you are to Facebook. The assumption is that you might not want everyone to know you’re at the hospital. So you could have your phone do one of three things when it sees that you’re at the hospital:

  • Block the feed outright;
  • Have your phone report that you’re at work;
  • Have your phone replay yesterday’s stream (straight out of a spy movie!)

Interesting possibilities. But this is where it got me: this is great – for this one scenario. But what about LinkedIn and Twitter and whatever else? And what about other scenarios?

For instance, if I’m traveling and go to the hospital, then I might want to alert a few people automatically. Unless, that is, I’m in the Emergency Room due to a DUI accident of my causing. Especially if I’m in cuffs. Then I need to alert whomever has bail money for me.

Or perhaps, if I’m at the grocery store, I want the phone to continue as usual, unless I’m late to an engagement like a dinner party, in which case I’d like it to alert the host that I’m on my way (and probably picking up some wine). Unless I’m at the pharmacy department – they don’t need to know about that. Especially if I’m picking up STD meds.

The power is both boon and bane. You can define so many scenarios for so many different tools and social media channels and friends or associates and… and… Each one sounds plausible – exciting, even. But if you zoom out into low earth orbit, you wonder: how is anyone ever going to manage all of the immense number of choices available?

It would be absurd to think that anyone would get a new phone and then start manually entering in their rules. It would take days. No one would ever do it.

So what’s the alternative? Perhaps use the automatic rules until they screw up, and then create an exception: “Here’s what I actually wanted.” The problem is, first of all, the damage may already be done: your phone may have dutifully sent your spouse a message about your stop for meds. Yeah, you can fix that for next time, but you really don’t want a next time.

Second, will people really stop what they’re doing and enter an exception? If it’s rare, it might not be that much of a burden. But if it’s rare, then a) a user is likely to think, “This is a one-off – I don’t want to spend time entering a rule I’ll never use again,” and b) unless the user interface is really good, the user will have forgotten (or never learned) how to enter a rule or select an option. Unless they’re sitting bored waiting for food at a restaurant, it’s likely they’ll simply blow it off.

If it’s not rare, on the other hand, then you will be beset by exceptions until you achieve an asymptote – that is, you’ve finally gotten enough rules entered that the system is now working more or less smoothly for you.

In both cases, there’s a good chance that users will give up on the feature and turn it off. (Assuming phone makers don’t remove that option.)

And none of this even takes into account the fact that we will all have different settings. Not the value of a setting, but the existence of a setting itself. Someone who got kicked off of Facebook for not using a real name might have a new account on Ello. The “powers that be” may not even be aware of Ello; at what point would it show up and have settings available?

Or what if you participate in some obscure sport? Would settings even be available for that? The whole settings “ontology” would have to be customized per user. Doable? Perhaps. Straightforward? I can’t imagine it would be.

I’ve seen many individual scenarios discussed; I have yet to see anyone take on the big-picture question of how users will be able to manage the extraordinary complexity and intricacy of keeping their devices aligned with their interests.

Of course, that I haven’t seen it doesn’t mean it hasn’t happened; it just means I haven’t seen it. So this is simply me raising the question for anyone to answer: Is context awareness manageable by the average user?

Tags :    1 comment  
Dec 08, 2014

On the Scene: Almost-Instant Semi Quotes with eSilicon

posted by Amelia Dalton

Historically, it has taken months, possibly many months to get a quote for semiconductor production. Lots of paper was involved - maybe a fax machine or two. Getting a MPW quote (in the old days) could be more like dentistry (get it? pulling teeth??!) than semiconductor design. Those days could be long gone if eSilicon has its way. In this episode of On The Scene, I chat with Mike Gianfagna and Bill Isaacson from eSilicon and get a MPW quote completely online and in less than three minutes. Take that fax machine!

Tags :    0 comments  
Get this feed  

Login Required

In order to view this resource, you must log in to our site. Please sign in now.

If you don't already have an acount with us, registering is free and quick. Register now.

Sign In    Register