What do I mean by AI being a fake thing? That it adds a layer of religious thinking to what otherwise should be a technical field. Now, if we talk about the particular technical challenges that AI researchers might be interested in, we end up with something that sounds a little duller and makes a lot more sense.” (Jaron Lanier)

“The other problem with AI dystopias is that they project a parochial alpha-male psychology onto the concept of intelligence. Even if we did have superhumanly intelligent robots, why would they want to depose their masters, massacre bystanders, or take over the world? Intelligence is the ability to deploy novel means to attain a goal, but the goals are extraneous to the intelligence itself: being smart is not the same as wanting something.” (Steven Pinker)

Quotes from “The Myth of AI”, Introduction by John Brockman, on Edge.org: https://www.edge.org/conversation/jaron_lanier-the-myth-of-ai


IMG_20160324_000625

This year, parking in the Hilton garage was only $20 per day, cheaper than expected. Friday started annoyingly, with an inflated breakfast price and sidewalks clogged by cellphone walkers. But this was SXSW Interactive after all, so the problem was all mine.

I attended panels and presentations about AI (or “cognitive computing” as IBM more modestly calls it.) AI was one of the core themes at SXSW this year, and a good match to my own interest. I had recently read The Cambridge Handbook of Artificial Intelligence and Twelve Tomorrows: Visionary stories of the near future inspired by today’s technologies (edited by SXSW in-house futurist Bruce Sterling.) I recommend both.

Since then, I’d been pondering how to organize my thoughts about AI. Alas, today my filter bubble fed me the Edge.org Reality Club article quoted above, where greater thinkers offer well-written insights on the topic. The article is from 2014, but holds up. So, I refer to the experts.

There are two things on my mind, though. The first is around design. AI is anthropomorphized, giving it relatable character: Siri and Cortana have distinctive voices. Watson has a distinctive voice and a swirly “brain-planet” like icon. AI entrepreneurs show us fake simulation-like videos where data runs through algorithms and forms brain-shaped 3d plots. These things are done as product branding, but also just to provide the general public with a mnemonic anchor to relate to. The actual technology is fairly hard to explain, after all. An intrinsic motivation for this design approach is to give the impression of human-ness: “AI is like humans, and will be as smart as humans soon.” This, however, is not actually true, and misguiding, even dangerous to the benefits that AI could provide. See the article above. Making AI appear human-like is an obvious design direction to take, but ultimately the wrong one, like a diskette icon for “Save.”

Finding the right approach is a major design challenge. AI provides conversational UX in many cases, and we converse with persons. We’ll find conventions for person-like automatons. We’ll find ways to highlight the increasingly comprehensive filtered nature of information presented to us.

The second point to consider is the impact on employment. Small town economies sustained by truck drivers won’t know what hit them. Many white collar jobs today will become infused with some level of data-handling, from engineering to K-12 teaching, reducing the number of people needed to carry the workload.

As pointed out by many proponents, new jobs, unknown today, will be created. After all, the web killed lots of jobs in the traditional publishing field, but also created entire new web-this-or-that industries. And today, the Sharing economy allows everyone so inclined (or forced to) to squeeze extra value from their (depreciating) assets.

As with all technological employment impacts, people holding lower-skilled who are not in a position to just pick up one of the shiny new jobs are most affected. The ranks of left-behinds will swell. AI as the superhuman power, foisted upon America by the “tech class”, will not look good. Religion-like technophilia (see Lanier’s thoughts above) will get a massive amount of blowback from people who are worried about putting dinner on the table, who are mad about it, armed, and registered to vote.

Rising tide declarations from entrepreneurs won’t fix this.  The AI field could tune down the pseudo-human rhetoric and focus on highlighting beneficial applications. Characterizing the technology as a set of tools would steer the debate away from the non-issue AI robocalypse. In the meantime, it’s time to reconsider the old puritan trope that only a hard-working person is a good person / American. Maybe legal weed would ease tempers, too.

So much for a bit of soapbox. Back into the sandbox. I am looking forward to buying a car using AI instead of dealing with the guy at the dealership, to get the knowledge of the world’s best doctors connected to my vital signs, and to learning new things using a teaching system that tailors lessons to my understanding. Hopefully, then I’ll get past the first couple of tutorials for TensorFlow. Seriously though: the opportunities for education are incredible, and the 19th/ 20th century education system is ready for change on all levels. SXSW Edu (attended by Dr. Starbuck, my wife) apparently is the cool conference today, not SXSW Interactive.

Over the last year, I have been spending some time modeling sculptures and 3D printing them. This included some utilitarian parts for my racing drones, some photogrammetry-modeled classical sculptures (more on that another time), and some original sculptures. I’ve been pursuing simple expressive shapes, trying to avoid the 3d printing tropes that make everything look like clever, twisted meshes.

OLYMPUS DIGITAL CAMERA

Kato is a result of that: a cat made from five ovoids, with character. After the initial prototypes in PLA and similar materials, I got nice material quality results with a woodfill PLA which makes the print feel like balsa. Biofila Linen, based on lignin, was another great choice, creating a hard print that does not feel like a plastic. I then uploaded the model to Shapeways, for printing in porcelain clay, which was then glazed and fired. The result is something entirely different from a home 3d print: a little sculpture. A heavy, solid, refined object.

The metal materials offered on Shapeways are interesting too. Those require a hollow model though, to keep things (relatively) affordable. I’m ordering some prototypes.

OLYMPUS DIGITAL CAMERA

Matte black, glossy white & red porcelain.

Kato can be ordered from my Shapeways store.

Regarding Shapeways

I have been using Shapeways’ printing services since 2008 or so for various projects. One uploads a model, picks a material, and receives a print in the mail, billed by the cubic centimeter of material, from plastic to platinum. Shapeways offers print material options and quality that can’t be matched by maker-level 3d printers. There also is an active community, and one can set up a store. What Shapeways does not have is curation. Everyone can upload and sell anything, as long as it prints well. This is, of course, the right thing to do. As a side effect, the art-related categories are loaded with thrash – I mean to say: contain some things that are liked by people other than me.

Regarding Affordable 3D Printers

…In case you are considering getting one: I bought a used Makerfarm Prusa i3v on Craigslist. It was a great deal on a well-designed open source printer. The thing to understand about open source maker-level 3d printers is the following: they are not appliances. They demand a relationship. Using such a machine means experimenting, troubleshooting, tuning and upgrading software and hardware. A well-tuned cheap printer rewards the owner with great results (which hold their own against systems that cost five times as much) – until it doesn’t, and then circle starts anew. It’s a form of fun.

Some thoughts about what it takes to get started with FPV racing (“Drone Racing”.) Lately, the topic has been coming up in conversations regularly. FPV racing, and free-flying a racing drone is a compelling, fun experience on many levels. The visceral impact is far stronger than flying a regular “aerial robot” type camera drone. The systems are so robust that crashing is not a big deal, so pilots find themselves free to push their skills without worrying about losses.

Besides educating yourself online, come visit a local race. Racers welcome people interested in the field.

Disclaimers

This is not a mainstream past time. It takes more than just buying some things and plugging them in. Thus the post is not a simple shopping list. Once the FPV racing topic becomes so commodified that anyone can do it without much thinking, you might find that today’s practitioners will have moved on to gardening or carpentry.

The field is changing quickly, so this post won’t age well. If you come reading this more than a few months after the posting date, then beware and do your own research.

I make specific recommendations for brands here. These are based on my personal opinions. Your mileage may vary.

IMG_20160207_154537355

QAV250

Where and How to Fly

When learning with a small quad (see below), practice at home in your yard, or even indoors, if the system is small enough. Or find a secluded spot in a park, without dogs and children running for your quad. Always avoid flying over people, traffic, and other people’s back yards. Don’t be “that guy”, doing creepy and reckless things with a drone.

Consider getting an AMA (Academy of Model Aeronautics) membership, which provides liability insurance. Most home owner’s policies specifically exclude flying things from coverage.

Once you fly an actual racing quad, you’ll need some open, empty space, since you’ll be covering a lot of ground fast. Again, consider parks, or open country, and act like a responsible adult. For FPV flying, bring a spotter: someone who stands next to you and provides you with situational awareness while you are wearing FPV goggles.

Actual races are often organized at RC club’s flying fields. Here, AMA membership is mandatory. The MultiGP organization has local chapters all over the place to organize races, under a common rule set and scoring system. There are other organizations too.

Learn With A Small Training Quad

To get started with flying quadcopter in the first place, get a small, cheap one to train with. I usually recommend anything under $250 (copter, transmitter, batteries, charger in the box) from the Horizon Hobby Blade brand. Get a “hobby grade” system (Blade systems are), not a super-cheap toy from Fry’s, so you can get replacement parts. Buy online or locally at a hobby shop. Hobbytown USA is the hobby shop here in Austin, with two locations.

You’ll want to learn line-of-sight flight first, looking at the quad. You can get to FPV (looking through the on-board camera) later. The LOS reflexes learned on a cheap, safe system, flying in your yard or in your living room, will bring safety and peace of mind later.

Some of the little Blade copters come with FPV options. Fun to have, but not necessary to start with. If you get one with an FPV camera, you’ll need to get FPV googles too, which are a separate expense.

IMG_20160207_161018219

Small training quad with FPV, and full-featured RC transmitter.

Flight Modes and Simulator

The small quads will all have stabilized, “assisted” flight modes, where the quad auto-levels: when you let go of the sticks, the quad goes horizontal and just sits there (or scoots around.) Any quadcopter has those flight modes, and they make keeping the quad in the air relatively easy.

A source of fun in FPV racing comes from the fact that races are flown in “Rate Mode”, where you manually tilt the quad forward, backward, and sideways, and it just goes where you point it, and it does not recover to a level attitude on its own. Tilt forward to gain speed, tilt backward to slow down. Tilt sideways and backwards to turn. Like an airplane. Once learned, this way of flying becomes second nature and brings a level of connection between pilot and machine that can’t be matched by any assisted mode.

It’s a bit intimidating to learn, though. Thus, use a simulator. FPV Freerider is good and cheap. Get a cheap controller that mimics an RC transmitter for it. Don’t bother learning the simulator with mouse and keyboard or a gamepad – you want to learn the right muscle memory. If you are ready for the plunge, you can also buy an actual RC transmitter that hooks up to the simulator, the Taranis. See below.

The simulator will also allow you to get comfortable with the “through the camera” FPV perspective. Crash a few hundred times in the sim, and you’ll be running through race gates soon.

Components of a Racing System

To actually do racing, you’ll need a quadcopter, a transmitter, a set of FPV goggles, some batteries, a charger, and replacement props. You’ll also want an HD camera. System cost can quickly get to a thousand dollars or more. This investment yields many hours of enjoyment.

When making spending trade-offs, consider that your radio and your goggles will last a long time, but you’ll be beating up your copter.

How it Works

An RC radio is used for controlling flight. RC radios today use spread-spectrum frequency hopping, so interference between different pilots’ radios is a non-issue.

The actual flying is not that different from flying an RC plane. A flight computer on the copter takes the inputs from the RC receiver and computes what the four motors should do (speed up or slow down) a few hundred times a second. Per motor, an ESC (electronic speed controller) translates the control commands from the flight computer into power fed from the battery to actually move the motor. The propeller attached to the motor generates some amount of lift, which depends on the speed of the motor. The attitude and position of the copter changes as a result of the lift. The flight computer’s sensors sense a change and compute a new position and attitude. The flight computer compares this data to it’s own model of where the copter should be at that moment (based on current flight mode and RC radio control inputs) and generates new commands for the ESCs.

A separate radio system transmits analog video from the camera on the copter to the pilot’s FPV goggles. The video is transmitted on a single channel, and the pilot has to match the transmitter’s channel (on the copter) to the receiver’s channel (in the goggles) and ensure there is no interference with other pilot’s video systems on the same or nearby channels. Video transmission is a bit primitive, and cooperation between pilots is required.

On some systems, an OSD (on-screen display), a small computer of its own, translates data from the flight computer into a read-out that is superimposed on the video image, showing things like power usage, a flight timer, or an artificial horizon.

Some goggles have a DVR built in that will record what you see during the FPV flight (a low-quality video.) Many pilots strap an HD camera to their copter to make high-quality recordings of the flights.

A Hands-On Experience

Be prepared to occasionally make repairs that involve screwdrivers and soldering, and to do firmware configuration. None of this is actually hard. Most of it is satisfying work. Youtube is your friend.

But do be aware that FPV racing is not a high-convenience “go to the store and push a button” consumer experience. FPV racers are helpful people, and can get you flying again, helping out with parts and expertise, but you still want to develop your own repair skills.

Racing Quadcopter Choices

While a number of size classes exists (see the MultiGP website), in practice everyone currently flies 250-class quads and smaller ones.

Size Classes

250-class means that the diagonal distance between motor axes is 25cm. This allows for 5 inch props. Just a few months ago, 250 was the only size flown. Today, a number of 210-size airframes, still allowing 5 inch props, are very popular, as are 170- and 180- size systems, which fly on 4 inch props. 5 inch bring more performance than 4 inch, but smaller systems are lighter and more nimble.

Some folks are even moving down to 150- and 120-size systems with 3 inch props, trying to beat the FAA registration weight limit. At the point of this writing, it’s not clear to me if this is a trend with momentum.

Frame Shapes

Many 250 quads have “H” shape airframes, with an elongated body holding the components. Many new, smaller airframes are “X” shaped, with a small pod in the center holding the components in a stack.

X airframes ensure that the weight of the system is all in the center, creating an airframe that is balanced and as nimble as it can be. “H” shaped airframes have more mass off-center (to the front and the back of the hull), so they can be less nimble, since the off-center mass pulls sideways during turns.

“H” shaped airframes offer more room for components, though, which makes them easier to work with. When the battery is mounted in the center, on top, an “H” airframe can also achieve decently-centered mass and balance. When the battery is mounted in the back (this used to be a common practice), then the battery acts as a large off-center mass.

IMG_20160207_154512324

250-size QAV250 with 5 inch props, center-mounted battery. 180-size Morphite V2 with 4 inch props. Both are “H” frames and both use bullnose props.

Flight Controller

Besides the physical attributes of the system, the flight controller, its firmware, and the firmware’s “tune” have a great impact on flight characteristics. The standard flight controller is the Naze, an open source board, and its derivatives. The most popular firmwares for it are Cleanflight, and the race-optimized Betaflight, a fork of Cleanflight.

These firmwares offer different algorithms for actually computing the system behavior. Of those, the Betaflight Luxfloat algorithm is getting a lot of attention currently, but there are a number of good choices, which, in practice, offer either a “smooth” or a more “robotic” feel to the pilot. They all fly well, they are a question of taste, and they can be changed easily.

The flight controller’s algorithm requires tuning: adapting a number of parameters that determine how the PID loop in the algorithm acts, determining how the system will fly. Tuning is a matter of changing a value, testing the results, and then making the next change. The goal is a “locked in” feeling, meaning the copter does what the pilot wants it to do, in a intuitive way. Pre-built systems usually come with a tune. Homemade systems need to be tuned. Every algorithm needs its own tune.

Build or Buy

The best deal can be had by building a system from scratch, sourcing all the parts and putting them together. Building takes between a few and many evenings. Many build tutorials are out there. Component choices vary in quality. Some folks have fun with cheap ripoff airframes (the ZMR family) and cheap motors from Hobbyking or Banggood.

Sourcing from US-based vendors costs more. While most components still come from China, the quality control is usually better, and sometimes the workmanship and the materials are too. For example, a super-cheap carbon fiber airframe is likely to use a low-density carbon weave, and even just use fiberglass filling between outer carbon fiber layers. An expensive carbon fiber airframe might use mil-spec carbon fiber with a high-density weave.  Your $20 or $120 investment then means a broken airframe in a crash vs. an unscathed one.

Examples for good US-based vendors are: GetFPV, ReadyMadeRC, BuddyRC, QuadQuestions, Armattan. There is an entire cottage industry of these, offering many great airframes: Lumenier QAV250, -QAV210, -QAV180, -QAV-R, Shendrones Tweaker, -Krieger, Armattan-Anything, Blackbolt XBR220, QQ190, QQ Sparrow, RMRC Hellbender.

Online Instructions

Build instructions can be found on Oscar Liang’s blog, on FPV-Flightclub, and on RC Model Reviews for example. These sites are great sources of knowledge. Another source of knowledge are the forums on RCgroups, but threads there can balloon to thousands of posts, making them useless.

Besides building instructions, also look for tuning instructions. The flight controller requires tuning for it’s PID loops, which control how the system flies. PID tuning is dependent on your specific setup (airframe, motors, weight, props and so on) and on the firmware and algorithm used.

Pre-Built Systems

Some vendors also offer ready-to-fly systems, or systems that are only missing the video components and receiver (GetFPV, Armattan and QuadQuestions come to mind.)

A few systems, namely the variations of the ImmersionRC Vortex, are factory-made “ready to bind” – just bind to your transmitter, or install your own receiver.

Some systems are available fully built and with a radio. The Walkera Runner, one such system, is not very good. The Graupner Alpha is very good, but as a German vendor, Graupner has almost no presence in the USA. The TeamBlackSheep Gemini is actually a hexacopter, but available in a fully built and complete package.

Buying Used

Besides building or buying pre-build, buying a used system can be a good choice. Many fliers have more copters than they need, and are always building the latest and greatest one. Gently used, well built systems are often available from local fliers or on the multi rotor for sale forum on RCgroups. Buying locally will be the simpler choice.

Radio System Choices

Radio systems are vendor-specific: a particular vendor’s transmitter usually only works with that vendor’s receivers. “Usually,” because there are some modular systems, where a transmitter can accept an RF module (that does the actual radio broadcasting) which is compatible with another vendor’s receivers.

All radio systems available today, cheap and expensive, do work OK, and don’t interfere with each other. Almost all racers fly 2.4gHz systems, and almost all radios are 2.4sHz systems.

The go-to radio for FPV racing is the FrSky Taranis. It’s a cheap-ish system, of low hardware quality, that copies features from high-end systems and uses a powerful open source firmware. Almost everyone uses it.

Second are the Spektrum radios, like the DX9. They are more expensive on a per-feature basis, closed-source, and offer somewhat better build quality. They have the advantage of working with off-the-shelf multicopters and rc planes from Horizon Hobby, where one can buy a “Bind and Fly” machine that will easily be paired with one’s Spektrum TX. This extends to the small Blade-brand quadcopters that I recommend for initial learning. When getting a small Blade quad, consider buying the “Bind and Fly” version and buying a fully-featured transmitter at the same time, instead of buying the “Ready to Fly” version that usually comes with a low-feature transmitter in the box. The full-feature transmitter will fly your racing quad too. The full-feature transmitter can be a Spectrum transmitter, or it can be a Taranis with a Spektrum RF module.

Many other good RC radio brands exist, like Futaba and JR, but they have no presence in FPV racing. The most awesome high end radios are made by Jeti Model.

FPV Goggle Choices

You wear the FPV goggles and look at their screens to see what the camera on the quadcopter sees. Goggles vary in image quality and field of view. The bigger the field of view, the better: it’s more immediately immersive. After using goggles with a small field of view for a while, they become very immersive too, but the brain takes a while to transition from “I’m looking at a tiny video screen” to “this is the world now.”

Most goggles available offer one small LCD screen per eye, plus optics in front of each screen. The optics quality varies. Glass optics are better than plastic optics.

Many googles offer IPD adjustment, to make the distance between the screens match the distance between your eyes. This becomes more important with larger fields of view, as the optics for large-FOV systems are less tolerant to eye position variations.

The most popular goggle brand is Fatshark, and their best goggles are the Dominator 2. (Or the even more expensive Dominator 2 HD.) Also very popular is Boscam.

IMG_20160207_155634442

Dominator HD V2 goggles

Besides the LCD-per-eye goggles, there is a type that can be described as “Screen in a Box”, where a single LCD screen produces the image for both eyes to see. Examples are the Headplay HD, and the Hobbyking Quanum. In this type of device, the screen is the size of a phone screen or small tablet screen, and a fresnel lens in front of it provides magnification. These single-screen headsets (“goggles” does not apply) are very large compared to normal goggles, but still quite light. They offer enormous fields of view and are cheaper than normal goggles. Their drawback is that they are harder to look at. There is a lot of cognitive processing involved in staring through two sides of a fresnel lens, off-center, at a big screen very close to the eyes. As a result, some young FPV fliers love the single-screen headsets, whereas some older fliers (like myself) get headaches and lose orientation when wearing them. Traditional goggles are optically engineered to avoid these issues.

12003216_10153070710821937_818358170145039560_n

Headplay HD headset

Lastly, there is a new type of headset, introduced at CES 2016 and just now arriving on the market, doing retinal projection. No traditional screen is used at all, but rather a tiny DLP projector per eye, projecting an image directly on the user’s retina. The Glyph is the first headset of this kind. Check the reviews as they start appearing. The Glyph is primarily marketed as a virtual reality headset, but used by the pilots of the DRL.

Almost all available googles have a video receiver built in. If a particular set of goggles has no receiver built in, an external receiver is needed.

Different vendors use different frequencies for their video transmission. Specifically, Fatshark and Boscam use different bands. Many receivers can only receive their own brand’s bands. Luckily, most current video transmitters offer many transmission bands, allowing them to work with most receivers.

Instead of using goggles, one can also fly FPV while looking at an image on a monitor, using an external receiver, without wearing goggles. This is less immersive and less fun, and not practical for racing, where immersion is required for quick reactions for “locked in” flying. The alternative is crashing.

Visit a local race to try on different goggles and headsets, to see what works for you.

Batteries & Charger

Like most laptops and mobile phones, quadcopters use LiPo (LIthium Polymer) batteries, because this battery technology provides the best power-to-weight ratio.

LiPo Safety

These batteries must be handled safely. When damaged or mistreated, a LiPo battery can and will ignite, with a very hot flame that quickly spreads if combustible materials are present. When LiPos first came out, many RC pilots’ cars and even homes burned down in LiPo fires, and it still happens today. A fire can be a result of mechanical damage or of thermal runaway from drawing to much current, from over-discharging a battery, or from shorting out a battery. Fires can start immediately when damage occurs or 10 or 20 minutes later.

Battery Sizes

250-size quads typically use 1300 mAh batteries, and smaller quads usually use 1000 mAh batteries. Propulsion systems are usually dimensioned so that 3s (3 cells in series) batteries provide “good” power, and 4s batteries provide “crazy” power. Most setups are quite tolerant, handling lots of combinations of prop sizes and batteries. But when too much power is used, a motor can overheat or an ESC can burn out. Some systems are only designed to handle 3s batteries.

For FPV racing, a battery must have a high C-rating. The C-rating describes how much power a battery can deliver when asked. For example, a 1000 mAh battery with a 20c rating can deliver 20 A power, and a 1300 mAh battery with the same 20c rating can deliver 26 A. If the C-rating is lower than the power required by the propulsion system, then the battery will fail to provide enough power, and will eventually overheat. For example, using the 1300-20c battery with a copter where each motor draws 12 A at full throttle, resulting in a 48 A peak demand, would be a bad idea.

For a given capacity, batteries with higher C-ratings are larger and heavier than those with lower ratings. Having C-rating “overhead”, using a battery with a higher rating than needed, is still advisable, as it will allow the battery to operate well within its limits, instead of being “pushed.” The battery will run cooler.

Many battery brands make optimistic claims about their C-ratings. Proven, reliable batteries are Tattu 75c, and, to a lesser degree, Hobbyking Nanotech.

Charging

LiPo batteries require LiPo chargers, which use balance leads to bring different cells in the battery to the same voltage. Batteries can be charged at 1c rate, taking an hour to charge. Higher charging rates can also be used for many batteries, pumping power into the battery faster, resulting in a quicker charge, but creating more thermal stress along the way. For long lasting batteries, charging at 1c is best.

When storing LiPo batteries, it is best to have them charged to only about 2/3 capacity for long life. Fully charged batteries deteriorate over time.

Some chargers allow independent charging of multiple batteries at the same time. This is useful when going through many batteries during a flying session.

A device called a “Paraboard” can be used to hook up several batteries to a single-battery charger at the same time. The Paraboard distributes power between all the batteries, while the charger thinks it is only charging a single battery. This works best when all the batteries are at the same charge level. Otherwise, cross-currents can cause damage. Paraboards are convenient, but somewhat risky to use.

HD Camera

GoPro cameras are a common choice to make HD flight recordings. They are a bit bulky and heavy though, not to mention expensive. A range of smaller, cheaper HD cameras is often used instead, such as the Xiaomi Yi, RunCam HD2, Mobius, or the Foxeer Legend. The HD camera is not connected to the actual FPV video system.

midair

The moment before impact, as caught on a GoPro

Replacement Props, Parts and Tools

Crashes, or sporty landings, are common in FPV racing. Mid-air collisions with other copters also happen. Racing quads are built to last, and they take a lot more abuse than any other RC multicopter. Not fearing expensive repairs is part of the appeal of flying a racing quad. But we also push our racing quads hard. Rarely, a crash will break a part of the copter’s frame, requiring installation of a replacement part.

It is quite common to break propellers (props) though. Installing new props after every crash gets old. Some props are marketed as “unbreakable.” These do indeed break a lot less often, and are thus widely used.

IMG_20151205_120539934

Hard landing

When choosing props, steeper angles (5-5 instead of 5-3, for example) provide more thrust, if the motors and ESCs can handle them. More thrust means more speed for racing, but makes the copter harder to control. Less thrust means more stability.

Bullnose props provide more thrust than conventional props, but use a lot more power, operating at lower efficiency. A bullnose prop is shaped like a prop of a larger size, but cut down to a smaller size. The blade cross-section at the outer edge of a bullnose prop is wide, whereas a regular prop has a tapered edge.

A 3-blade prop provides more thrust than a 2-blade prop, but also draws a lot more power.

Generally, achieving maximum speed conflicts with achieving maximum flight duration. The sweet spot of a propulsion system, providing the longest flight times, uses a prop size and angle lower than the maximum thrust setup. For racing, speed can be more important than efficiency, since even on low efficiency, one usually gets enough flight time to fly a few laps.

Repairs & Tools

Sooner or later, something needs to be soldered. Sometimes, a vendor will have the exact cable with the specific plugs that you need, but chances are you’ll have to make your own. Especially when doing anything with the video system (like installing a new video transmitter), the need for custom wiring quickly arises. You’ll also want to shorten cables to just the right length when installing something, so there are no “wire nests” clogging up the limited space in the airframe.

Have a temperature-controlled soldering station with a fine tip available, as well as basic wiring tools (clippers, pliers, strippers.) A “third hand” (flexible, posable clamps) or a wiring jig can be very useful. Small hex- philips- and flat screwdrivers are also good to have. In the field, a ratchet with a bit sized for the prop nuts is useful, making prop changing easier than using pliers and fingers.

Happy flying.

IMG_20160122_191353123

On the bench, to soon fly again

I just spent 5 quarters designing a new experience for a legacy enterprise change management solution. Along the way, I learned a lot about the topic, both from documentation and from sparring with the experienced project team. My design proposals often served as vehicles for additional requirements discovery: I’d think I’d have a comprehensive solution for a topic, but team members would point out additional conditions to be addressed. We would then have a long conversation, in which development team members aligned each other’s diverging interpretations of the topic at hand, and educated me on the nuances. To achieve an actual reduction of complexity In some cases, we could “dismantle” requirements that turned out to be mere assumptions. In other cases, there clearly were requirements, and the design had to embrace the complexity. Which is “what we do” in enterprise design.

OLYMPUS DIGITAL CAMERA

Unrelated Image

This makes me, more or less, an “expert” on the topic of change management. I use quotes around “expert”, because, compared to the team members who spent an entire decade working on change management, I am clearly not an expert. As I have been watching my grasp on the topic grow, I have also seen how limited my understanding is.

But my now-sixteen month tenure on the project left an impression. Based on my learning, I’ve been envisioning what a future of change management could look like. In more general terms, the challenge is how to design for process-oriented experiences.

 

Current Design

Well, I’ve just done a design, so isn’t the answer in there ?

Naturally, I believe in the solution I worked on with the team. It’s a web application, following inductive principles, using the proven “context within a context” approach. The approach is to use the application structure and flow to guide and reinforce the user’s understanding of the underlying object structure, dependencies and sequences. So far, so good. Feedback is positive all around.

 

Reliance on the User

There is an implied expectation to make this type of design work: the user is meant to learn the application, to engage and get a handle on the structure, dependencies, and sequences. “Work with me here,” the design asks the user. “Here is an elaborate, custom structure. Explore it, and you will understand,” the design tells the user. As an immediate answer, users are happy to do just that, if the experience offered is engaging in a “fierce fun” game-theory way. But still, the user has to engage with the design. In a general way, this is how software design has been operating. Especially for business software, “learn the tool, it’s a good tool” is a reasonable ask.

 

Why Make the User do the Work ?

Take a step back. Is there a different way of looking at this?

The user has goals, and software tools help with achieving these goals. The user has a life and a workday outside a given specific software tool, where many goals may be competing for attention.

In this context, a software tool can become a better fit for a user’s life by becoming less demanding: ask for less of the user’s time and attention. The software tool should offer meaningful choices to advance the user’s goals, while allowing the user to carry as little contextual overhead as possible. The tool should be as little a distraction as possible. A “classical” software experience, providing the user with paths through hierarchical screens, even when perfectly designed, does not reach that ideal.  

Being less demanding means presenting “just in time” choices when needed, with just the right amount of context, leading to the next choice in a “shallow” flow, until a desirable end state is achieved.

 

Semantics-Driven UX

A just-in-time UX approach places new kinds of demands on the user interface delivering the choices, and on the design process behind it. Instead of offering the user a framework of orientation to understand the current situation, the just-in-time user interface must have its own architectural semantic representation of this framework, to distill choices down to the right ones. In other words, the user’s cognitive workload shifts to system workload and state management, insofar as the system will have an internal coding of the choices and state sequences.

In a traditional hierarchical user experience, choices and state sequences are present implicitly. The same hypothetical application, designed hierarchically, or as a low-context step flow, ideally is Turing-equivalent between the two instances (or equivalent in choices made, expressed in an experience-centric measure that is equivalent to computational turing equivalence, but now we are getting geeky.)

 

Encoding Eliminates Hand-Crafting One-Offs

In a traditional user experience, choices and state sequences are expressed via a hand-crafted hierarchy and flows. This traditional process of “User Experience Carpentry” is fun and proven (the UX field has just been getting to a point where we can finally say that.) But all-manual user experience carpentry is also error prone and labor intensive. Patterns and component libraries act as efficiency band-aids with limited leverage.

 

Design as Transformation

To get a just-in-time experience, a semantic encoding must play hand in hand with a delivery mechanism for the experience, and offer a way to infuse this delivery with design intent. Sensible design choices in this context are just as important as before, but they change in nature: from hierarchy and flow decisions to transformation decisions. Think of DNA being transcribed by RNA to build proteins, with knobs to turn on the RNA.

 

Better Decisions

A “lighter,” more on-point user experience, will allow users to make better decisions with less cognitive workload and with less time investment. In other words, fewer mistakes will be made, more decisions can be made, and productivity raises. Deep hierarchical applications, even well-designed ones, will suddenly look very old.

 

New Forms of Conversation

An additional benefit of a just-in-time experience is that it does not have to be delivered in a traditional user experience consisting of screens, windows, dialogs, and widgets. It can, instead be a spoken dialog. Or a written dialog, in emails or chats. Or a hybrid. The point is that when a self-contained encoding of a given set of choices is available, it can be expressed and acted upon in many forms.

Lastly, the form of semantically-enabled UI will be a foundation for working with a different type of user: machines. It can enable a scalable approach to automation, allowing a software to walk through a user experience just like a human user would, without requiring a custom integration between the tool and the “user software.”

 

Going From Hard to Easy

The innovations envisioned here (sequential UX, automation, UX as conversations) are possible today and implemented today, and have been described before. It all can be done if the budget is right. Achieving those experiences simply requires a whole lot of UX carpentry, which is not that different in effort from making wireframes for an old-fashioned hierarchical web app. The point is though, that the semantically enabled UI, in principle, allows us to avoid costly one-off solutions.

 

Designers Alone Won’t Do This, But Designers Will Make This Succeed.

Moving user experience forward will not be achieved by designers alone: progress will depend on technical architecture. Besides the actual technical challenges involved, a key success factor will be to keep the resulting tools and conventions within designer’s reach. There is a fine line between tools that actually enable designers, and technologies that take design out of designer’s hands. The latter breed a specialist who can handle the tool, but then the tool fades into irrelevance when technology moves on. When the toolset and the underlying technology are decoupled, then either can evolve freely and absorb innovation, for better execution capabilities and for best-fit authoring tools for specific users.
Anyways, there are some things to do.

OLYMPUS DIGITAL CAMERA

Unrelated Image

Honoria Starbuck (my wife, as you might know) has an art show coming up: Flaneuse, at the soon-to-be-opened Spellerberg Projects in Lockhart, Texas:

honoria-flaneuse-small

Artwork by Honoria Starbuck

January 16 – March 12, 2015
By appointment

Artist’s reception
Saturday, January 16, 5pm

Spellerberg Projects
www.spellerbergprojects.com
103 South Main St.
Lockhart TX

Email info@spellerbergprojects.com

Spellerberg Projects is proud to present Flaneuse, a dynamic installation of small-format paintings by American artist Honoria Starbuck.

The works in this exhibition are 5 × 7 inch postcard-sized gestural paintings, produced in watercolor, pen & ink, pastel, collage, graphite and acrylics. They are the product of the artist as flaneuse, maneuvering through contemporary life and engagement with art history.

Starbuck’s inspiration flows from the earliest of mark-making gestures, prehistoric cave painting, through art history and up to present day practices such as digital painting via tablet computers. The pieces explore color, shape and the interaction of materials, with subjects ranging from observed nature, abstract forms and fashion.

Commercial use of Drones (link to article on DIYdrones)

I’m at SXSW Interactive, and I’m glad to see that the programming is not “all drones all the time.” There are some drone related panels, but no hype-overexposure. So the industry can just go do it’s jobs, without having to respond to outsized outside expectations.

Yesterday, the running about commercial use of drones came around. It’s great news. Trappy (Pirker) had been fined for recklessness. That’s somewhat different from the more common situation where the FAA issues cease-and-desists agains commercial operators. Schulman and Trappy argued that the FAA did not actually have authority to interfere with drone businesses, and their view, for now, prevailed. Of course, the FAA will not just accept the ruling, and they promptly have appealed. At least, there is a grey area now.

The FAA of course is still in charge of airspace in the USA, and in charge of “things not falling on our heads”. There is a published roadmap for integration of small drones into the national airspace, and there is a rule making process around using drones in progress. The draft of these rules was just delayed again, by another half year, but eventually it will become available.

Some may interpret the current court ruling as a manifestation of the latent conflict of interest between the drone industry (open up the skies) and the FAA (stay out of out skies, or be really careful if you come in.) Looking beyond the ruling, of course we’ll need rules. (Michael Toscano of the AUVSI, in reaction to the ruling, quickly affirmed as much.) Maybe the grey area that’s open now can be used to establish some reasonable practices, while drone flyers can pay their bills, on the right side of the law.

p.s. http://www.grafsys.com

One flight near Nauders, Austria, and one near Lieser, Germany. We spent a week in Nauders, with a sportube full of rc gliders that were put to good use. Nauders, in Austria near Italy an Switzerland offers a fairly mellow alpine flying site, that sometimes offers great lift. Just right for the casual slope flying flatlander. Up the road from Nauders is the Fiss-Serfaus-Ladis area, which offers more alpine slope flying, for another time.

Lieser, in Germany is where I grew up. So when visiting, the planes were exercised there too. It’s in a wine region that has been shifting from quantity to quality products over the last decade or so. This leaves some former vineyards as open fields today, allowing slope flying without the constant threat of impaling the plane on a vineyard post.

Since the Gopro camera acts like a spoiler on the fairly small Easyglider, the actual soaring performance seen in the videos is not great. But either site does offer good conditions at times. In Nauders, I had been able to take a glider up almost 700 meters, following another glider with an altimeter, and in Lieser, I had gotten a one hour no-power flight on the day of the video. Both without the camera-airbrake, though.

Image

Open class Fusion, Graupner Amigo IV, FVK Bandit, Multiplex Easyglider