Was just going to suggest that! I’ve only recently started using it but so far it has been excellent. One tip… the code generator seems to do a lot of cleaning and simplification. Just today I had some ugly equations that took up many pages of Jupyter output but the generated C++ code was only about 150 lines and didn’t look particularly inefficient. It seemed to me on inspection that it found repeated product and sum terms and properly computed them once and stored them in scratch variables.
That is not quite the whole story. They are making it available for anyone in Oslo, at least. [1]
I live in Oslo and have an appointment to get a booster tomorrow. I have multiple friends and family members in the healthcare sector, and they are all warning me that we are likely to see a significant spike in the winter. At the same time we are seeing more and more sign that each repeated infection has increased risks. [2]
That's not what the paper cited by Time says. It is a study in the US veteran population, whose health condition is much poorer than the general population. Secondly, the study is set up to pick up bad outcomes. Third, the people compared are those with 1 and 2 infections after the second infection, rather than seeing the effect of each infection. Fourth, the paper is pre-Omicron.
There is also far more than that. The press spun it quite a bit, in a different way.
Personally, all my family members and myself have had a contact with SARS-CoV-2 within the past months. It wasn't particularly pleasant, but certainly not a "deadly virus" (prior vaccination helps, but I'm not a person at risk anyway). It will be the first of many. I stopped caring after I got my second dose, and there's no way I will care now.
It might not be a deadly virus now but it certainly was in mid 2020. And things mutate. Flu is harmless as well, except for the years when it isn’t. Minimizing risk for others is a positive human trait.
Flu is not harmless. Not at all. It is a bad disease when symptomatic (30-40% cases may be asymptomatic). In particular for young children and the elderly.
Flu for me was way worse than a SARS-CoV-2 infection (think 40C fever).
SARS-CoV-2 was deadly for a specific part of an immuno naive population (60+). The age stratification was known since February 2020 (first Chinese paper on that from the Wuhan data). My country had zero excess deaths in the 0-50 range since the start of the pandemic.
EDIT: Anti-SARS-CoV-2 vaccines do not block transmission, certainly not now. And as thus, their protection is primarily for one's own benefit.
Malaria is not an infection by a coronavirus (I do know one person too, FTR, my grandfather during the Albanian campaign in WWII). And it is not age stratified like a SARS-CoV-2 infection. And there's no 95% seroprevalence like with SARS-CoV-2 in most regions that had surges.
I have had a love-hate relationship with YouTube for a long time. On the one hand, many channels put out great content that I want to watch. But the recommended videos feature is exceptional at hooking into my reptilian brain and getting me to waste a lot of time on junk that I later regret as a big waste of time.
I looked into a lot of alternative clients, including FreeTube, in an attempt to solve this. I use Tube Archivist[1], an open-source project that allows you to subscribe to and auto-download YouTube channels and playlists or download individual videos and watch them in a minimal web interface. It works fine on iOS as long as you set it up to download in an iOS-compatible format. I currently have it running in a Docker container on my Synology NAS.
Darnit, I already wrote the rest of the comment below before noticing the "iOS" part of your comment. These are all Firefox add-ons for the desktop. Still, I'll post them in case they're useful to anyone else.
I use the unhook add-on for that, it removes all recommended videos:
youtube-audio lets you disable the video-feed. Nice when you're using it to listen to what are basically podcasts and you don't want to waste bandwidth:
Probably Lie Theory, or at least the small subset of it that is relevant for robotics.
It has taken me a long time and a lot of work to understand it to the level I have, and I still have a lot more to learn. But it is a fascinating subject and a very elegant way to represent some very complex optimization problems.
How is Lie theory relevant to robotics and optimisation? I looked into Weyl groups a bit during my undergrad, but didn't really go into Lie theory in much depth.
Lie theory is, among other things, a very elegant way to parameterize an energy function involving rotations in 3D space. For example the translation and rotation of multiple camera frames that observe the same set of 3D points. This is the basis of how Visual Odometry, SLAM or 3D reconstruction problems are solved.
There are other cool things like representing a probability distribution over a 3D rotation with an multivariate gaussian vector interpreted as a so3 or se3 vector. Using this you can for implement a very elegant extended kalman filter estimating a 3D rotation. Which usually can be a tricky thing to do.
In robotics we pretty much only use the SO3, SE3 and SIM3 groups so my knowledge is limited to those groups.
The problem with quaternions for optimizations is that they are over-constrained. Meaning the norm of the 4 values must sum to 1 for it to be a valid 3D rotation. So you need to ether penalize a rotation with a norm different from 1 or constantly remap the quaternion back to a valid one in each iteration of the optimization. Neither solutions is very elegant.
A 3 valued so3 lie algebra vector is however always a valid 3D rotation, no matter what the values are. That makes it an ideal parameterization of the optimization problem.
The mapping from so3 lie algebra to SO3 lie group (which could be represented as a rotation matrix or quaternion) is a smooth differentiable function, which allows us to compute the jacobian we need for the gradient in the optimization.
It is way to hard and complex to do 2D and 3D visualization in C++. Toucan is my attempt to solve that. It can be called from anywhere with minimal code and gives you interactive 2D and 3D visualization.
I work in robotics and have for a long time been frustrated with how hard it is to visualize data in C++. I created Toucan to try to solve that. The project is still in a very early stage but has already started to become a useful tool.
The API still needs work but it’s getting there. Toucan can be called from anywhere in your code, and runs in its own thread to always remain interactive and responsive.
I haven't looked into this drone in depth, but I think it may be feasible. The DJI Mini2, which any civilian can buy for $449, has a similar size, weighs 249g, and advertises a flight time of 30 minutes. (In my experience with the Mini 1, it's plausible).
Side note: the DJI Mini 1 is an excellent drone -- stable flight, great camera, and great range. I wouldn't be surprised if it outperforms this 'Bug' drone by most metrics, except for the critical (at least for military purposes) "not being made by a Chinese company" metric.
It's hard to express how awesome the Mini is. IMO it's an underappreciated technological achievement at the once-in-a-hundred-years level. Only a few years ago, it was hard to get a decent digital camera for $400 and change. The Mini is a decent camera that also happens to fly.
When that realization hit me, it reminded me of the first time I saw a $29 DVD player for sale at WalMart. Some incredible things had to happen to make that possible, almost all of of them going completely unnoticed by almost everyone.
The BAE drone probably costs $50,000 apiece, and doesn't appear to be substantially different from the current-generation (or even the original) Mini.
That may be a little high. UAVTek's site says they beat the FLIR Black Hornet 3 bid[1]. Estimates based on US Army budget requests say the Black Hornet 3 is coming in between $15k and $16k per unit[2]. Probably still 5 figures though.
In addition you have to think of the trade-offs. If the army multiplied the price by 10 to get 10% better performance, that might save a soldier’s life.
Hmm, shades of Kipling‘s Frontier Arithmetic: “four thousand pounds of education/Drops to a ten-rupee jezail”....
That’s based on the value of the whole contract, which will include the costs of participating in trials, demonstrations, training, lifetime service and spare parts. There are likely to be a whole team of personnel dedicated full time to the contract for years, all included.
Apologies as I was ambiguous here. I find myself including both the GPS and flight controller computer/ carrier board in that 400$ price tag. I’m using Hex Cube Orange with Here gps units. This says nothing of cameras, radios, batteries, flight time, and live camera feed/control.
That's because of the energy density (per mass) of Li-Ion batteries. You basically hit a wall at 20 minutes that's very hard to push past (it almost feels like a law of physics).
But maybe they're using non-rechargeable batteries: zinc-air or lithium-iron-disulfide (ie. a 'lithium' AA battery) or something else? Trade offs are different for military.
Yes but there are no convenient charging points, if the battery runs out then you need to replace it, not charge it. All the more reason to throw used batteries away (they’re cheap and it’s less to carry).
I don't think it necessarily means non-rechargeable. Since charging usually takes a long time, generally in the field you carry a bunch of batteries charged ahead of time and swap them as needed.
The classic example of a good military-only battery would be those using molten salts as an electrolyte (typically at 150-500 ºC) [1]. Very high energy density (around 74 Ahr/kg [2] compared to LiPo's ~50 Ahr/kg [3]) and very high power density. Not finger friendly.
It actually appears both spellings are currently accepted in dictionaries, but in the fluid dynamics literature you will essentially only find "vortices", so that's a much better (more searchable) word to use.
They're claiming flight time - in the battery section of the infographic it claims "flight times of more than 40 minutes".
Tri-blade props are not the most efficient - we use them in racing applications because we want to generate maximum thrust and we don't need more than around 2-3 minutes per battery but we wouldn't use them in an aerial photo rig where we want long flight times - 2 blade props would be the go to. Also the pitch of those blades in the photo are quite aggressive.
I wonder if they have different configurations - as photographed I can easily believe the 50mph top speed with small but high kv (fast but inefficient) motors but I'd expect a different configuration for longer flight times.
From the photo it looks like the battery compartment is shaped for 2 cylinders, its about the right dimensions for 2x 18650 cells but it's impossible to say for sure. That would give up to 8000mah of 3.7v (1S2P configuration) or 4000mah at 7.4v (2s1p). They could be using different chemistry though, e.g. I usually fly LiHV cells which have a higher nominal voltage in the same pack size & weight - you do get very slightly more watt hours out of those packs (not as much as the numbers would suggest).
The given all up weight looks plausible based on the size of what's shown and the weight of 2 cells.
There's a few others, and in fact I considered doing this myself! But, I'd get bored of slowbelly 'flying' after about 15 minutes. I like to go fast.
I completely believe 40mins of flight time if you had actual resources to assign to this. Imagine using the 18650 as the frame and strapping motors to sticks, or other crazy ideas to bring weight down and time up.
In a military context, "flight" can mean "mission". I only suggest that something may have been lost in the language. I find that easier to believe than this thing hovering over a battlefield for 40 minutes.
Since you seem to have some expertise in this area, I am going to use the opportunity to ask a bit of a diagonal question I've wondered for a while:
I've seen claims about increase in airflow from bladeless fan designs (I know it's a bit of misnomer, as the blades are just hidden) for home applications. Is there a reason bladeless designs haven't been used in the UAV realm? I assumed the claims were just marketing, or the weight increase offsets the effeciency, or something like that, but thought I would ask.
I dont find it that unbelievable, there are a number of youtube videos of hobbyist grade quadcopters with what I believe are larger motors running 20-30 minutes on a pair of 18650 cells albeit without any autonomous capability.
It's all about the cube-square law, the same reason a butterfly can remain airborne for hours or days on a few milligrams of chemical energy.
Getting RC aircraft to fly in the more laborious, sedate manner of a full-size aircraft instead of like an insect fluttering through the wind is something of a challenge for scale model builders. When you build an airframe around tiny batteries as a design constraint, expectations from larger aircraft don't carry over well.
I agree the photos are probably just mockups. No way they would be using that sort of prop.
That said, Li-FES2 primary cells (at 400Wh/kg) have well over twice the energy density to LiPo/LCO rechargables (at 190 Wh/kg).
If I can make a custom 7" bi-blade quadcopter fly for 30 minute, 40 minutes is doable with better battery density.
In my own builds it is much easier to get long flight times with larger props but that is probably on account of the lesser need for exotic material science at larger scales.
The article makes it sound like they've got a whole bunch of these in active use by the Army (albeit for testing), so it'd seem a little odd to have a picture of a mockup. Maybe, I guess, but I'd have the assumption be that this is the real thing.
Given cost is not a problem, is it possible to get 40m flying time using exotic battery and ultra high strength/weight ratio materials?
I feel like the cost is deeply embedded in our psyche and people are comparing a military nano drone with DJI mini2 with entirely different engineering envelopes to work in.
DJI mini2 is made in thousands if not millions of quantities. There is so much cost optimizing. Even turning a part around is optimized because that’s more time to assemble.
