Neural networks are all the rage right now with increasing numbers of hackers, students, researchers, and businesses getting involved. The last resurgence was in the 80s and 90s, when there was little or no World Wide Web and few neural network tools. The current resurgence started around 2006. From a hacker’s perspective, what tools and other resources were available back then, what’s available now, and what should we expect for the future? For myself, a GPU on the Raspberry Pi would be nice.
We’ve been watching the development of the ESP32 chip for the last year, but honestly we’ve been a little bit cautious to throw all of our friendly ESP8266s away just yet. Earlier this month, Espressif released version 2.0 of their IoT Development Framework (ESP-IDF), and if you haven’t been following along, you’ve missed a lot.
We last took a serious look at the IDF when the chips were brand-new, and the framework was still taking its first baby steps. There was no support for such niceties as I2C and such at the time, but you could get both cores up and running and the thing connected to the network. We wanted to test out the power-save modes, but that wasn’t implemented yet either. In short, we were watching the construction of a firmware skyscraper from day one, and only the foundation had been poured.
But what a difference eight months make! Look through the GitHub changes log for the release, and it’s a totally new ballgame. Not only are their drivers for I2C, I2S, SPI, the DAC and ADCs, etc, but there are working examples and documentation for all of the above. Naturally, there are a ton of bugfixes as well, especially in the complex WiFi and Bluetooth Low Energy stacks. There’s still work left to do, naturally, but Espressif seems to think that the framework is now mature enough that they’ve opened up their security bug bounty program on the chip. Time to get hacking!
[Jason] has a Sonos home sound system, with a bunch of speakers connected via WiFi. [Jason] also has a universal remote designed and manufactured in a universe where WiFi doesn’t exist. The Sonos can not be controlled via infrared. There’s an obvious problem here, but luckily tiny Linux computers with WiFi cost $10, and IR receivers cost $2. The result is an IR to WiFi bridge to control all those ‘smart’ home audio solutions.
The only thing [Jason] needed to control his Sonos from a universal remote is an IR receiver and a Raspberry Pi Zero W. The circuit is simple – just connect the power and ground of the IR receiver to the Pi, and plug the third pin of the receiver into a GPIO pin. The new, fancy official Raspberry Pi Zero enclosure is perfect for this build, allowing a little IR-transparent piece of epoxy poking out of a hole designed for the Pi camera.
For the software, [Jason] turned to Node JS, and LIRC, a piece of software that decodes IR signals. With the GPIO pin defined, [Jason] set up the driver and used the Sonos HTTP API to send commands to his audio unit. There’s a lot of futzing about with text files for this build, but the results speak for themselves: [Jason] can now use a universal remote with everything in his home stereo now.
For most of us, a memory lapse is as harmless as forgetting to bring the garbage to the curb, or maybe as expensive as leaving a cell phone and cup of coffee on the roof of the car before driving off. But when the toddler sleeping peacefully in the car seat slips your mind in the parking lot, the results can be deadly.
We have no doubt that child detection systems will soon be standard equipment on cars, like backup cameras and trunk-escape levers are now. Not willing to wait, [ayavilevich] came up with his own car occupancy sensor for child seats (Update: We originally linked to the Instructable but [ayavilevich] wrote in and mentioned this is actual Hackaday Prize entry and he’s looking for more people to get involved in the project).
Dubbed Fochica, for “Forgotten Child in Car Alert,” the system is clearly a proof of concept right now, but it has potential. The Arduino Uno senses Junior’s presence in the car seat with a homebrew capacitive sensor under the padding of the seat and a magnetic reed switch in the chest harness buckle. An Android app on a smartphone pairs with a BLE module to get the sensors’ status, and when the phone goes out of Bluetooth range while the seat is occupied, the app sounds an alarm. Simple, but effective.
We like how well [ayavilevich] thought this through. Systems like this are best left uncomplicated, so any improvements he makes should probably concentrate on engineering a reliable, fieldable device. Another hack we’ve presented in the kid-safety space is fast stairwell lights for a visually impaired girl, which might provide some ideas.
Security for anything you connect to the internet is important. Think of these devices as doorways. They either allow access to services or provides services for someone else. Doorways need to be secure — you wouldn’t leave your door unlocked if you lived in the bad part of a busy city, would you? Every internet connection is the bad part of a busy city. The thing is, building hardware that is connected to the internet is the new hotness these days. So let’s walk through the basics you need to know to start thinking security with your projects.
If you have ever run a server and checked your logs you have probably noticed that there is a lot of automated traffic trying to gain access to your server on a nearly constant basis. An insecure device on a network doesn’t just compromise itself, it presents a risk to all other networked devices too.
The easiest way to secure a device is to turn it off, but lets presume you want it on. There are many things you can do to protect your IoT device. It may seem daunting to begin with but as you start becoming more security conscious things begin to click together a bit like a jigsaw and it becomes a lot easier.
Self-driving cars are, apparently, the next big thing. This thought is predicated on advancements in machine vision and cheaper, better sensors. For the machine vision part of the equation, Nvidia, Intel, and Google are putting out some interesting bits of hardware. The sensors, though? We’re going to need LIDAR, better distance sensors, more capable CAN bus dongles, and the equipment to tie it all together.
This is the cheapest LIDAR we’ve ever seen. The RPLIDAR is a new product from Seeed Studios, and it’s an affordable LIDAR for everyone. $400 USD gets you one module, and bizarrely $358 USD gets you two modules. Don’t ask questions — this price point was unheard of a mere five years ago.
Basically, this LIDAR unit is a spinning module connected to a motor via a belt. A laser range finder is hidden in the spinny bits and connected to a UART and USB interface through a slip ring. Mount this LIDAR unit on a robot, apply power, and the spinny bit does its thing at about 400-500 RPM. The tata that comes out includes distance (in millimeters), bearing (in units of degrees), quality of the measurement, and a start flag once every time the head makes a revolution. If you’ve never converted polar to cartesian coordinates, this is a great place to start.
Although self-driving cars and selfie drones are the future, this part is probably unsuitable for any project with sufficient mass or velocity. The scanning range of this LIDAR is only about 6 meters and insufficient for retrofitting a Toyota Camry with artificial intelligence. That said, this is a cheap LIDAR that opens the door to a lot of experimentation ranging from small robots to recreating that one Radiohead video.
When you think about which of the many technological advances of the 20th century had the most impact on the global economy, which one would you rank as the most important? Would it be the space program, which gave rise to advances in everything from communications satellites to advanced composite materials? Or would it be the related aerospace industry, which stitched the world together so tightly that you can be almost anywhere on the planet within 24 hours? Or perhaps it’s the Internet, the global platform for buying almost anything from almost anyone.
Those are all important, but for the most economically impactful technology of the 20th century, I’d posit that the lowly shipping container and the containerized cargo industry that grew around it win, hands down.