Need A Snack From Across Town? Send Spot!

[Dave Niewinski] clearly knows a thing or two about robots, judging from his YouTube channel. Usually the projects involve robot arms mounted on some sort of wheeled platform, but this time it’s the tune of some pretty famous yellow robot legs, in the shape of spot from Boston Dynamics. The premise is simple — tell the robot what snacks you want, entirely by voice command, and off he goes to fetch. But, we’re not talking about navigating to the fridge in the same room. We’re talking about trotting out the front door, down the street and crossing roads to visit favorite restaurant. Spot will order the snacks and bring them back, fully autonomously.

Spot’s depth cameras provide localized navigation and object avoidance information
Local AI vision system handles avoiding those pesky moving objects

There are multiple things going here, all of which are pretty big computational tasks. Firstly, there is no cloud-based voice control, ala Google voice or Alexa. The robot works on the premise of full autonomy, which means no internet connectivity for any aspect. All voice recognition, voice-to-text, and speech synthesis are performed locally using the NVIDIA Riva GPU-based AI speech SDK, running on the local NVIDIA Jetson AGX Orin carried on Spot’s back. A front-facing webcam supplies the audio feed for this. The voice recognition application listens for the wake phrase, then turns the snack order into text, for later replay when it gets to the destination. Navigation is taken care of with a Microstrain RTK GNSS module, which has all the needed robustness, such as dual antennas, and inertial fallback for those regions with a spotty signal. Navigation is no use out in the real world on its own, which is where Spot’s depth sensor cameras come in. These enable local obstacle avoidance, as per the usual spot behavior we’ve all seen before. But what about crossing the road without getting tens of thousands of dollars of someone else’s hardware crushed by a passing truck? Spot’s onboard streaming cameras are fed into the NVIDIA dash cam net AI platform which enables real-time recognition of moving obstacles such as cars, humans and anything else that might be wandering around and get in the way. All in all a cool project showing the future potential of AI in robotics for important tasks, like fetching me a beer when I most need it, even if it comes from the local corner shop.

We love robots around here. Robots can mow your lawn, navigate inside your house with a little help from invisible QR Codes, even help out with growing your food. The robot-assisted future long promised, may now be looking more like the present.

Continue reading “Need A Snack From Across Town? Send Spot!”

OpenMower: Open Source Robotic Lawn Mower With RTK GPS

Robotic mowers are becoming a common sight in some places, enabled by the cost of motors and the needed control electronics being much lower, thanks to the pace of modern engineering. But, in many cases, they still appear to be really rather dumb, little more than a jacked up bump-and-go with a spinning blade. [Clemens Elflein] has taken a cheap, dumb mower and given it a brain transplant based around a Raspberry Pi 4 paired up with a Raspberry Pi Pico for the real time control side of things. [Clemens] is calling this OpenMower, with the motivation to create an open source robot mower controller with support for GPS navigation, using RTK for extra precision.

The donor robot was a YardForce Classic 500, and after inspection of the control PCB, it looks like many other robot mower models are likely to use the same controller and thus be compatible with the openmower platform. A custom mainboard houses the Pi 4 and Pico, an ArduSimple RTK GPS module (giving a reported navigational accuracy of 1 cm,) as well as three BLDC motor drivers for the wheels and rotor. Everything is based on modules, plugging into the mainboard, reducing the complexity of the project significantly. For a cheap mower platform, the Yardforce unit has a good build quality, with connectors everywhere, making OpenMower a plug and play solution. Even the user interface on top of the mower was usable, with a custom PCB below presenting some push buttons at the appropriate positions.

OpenMower mainboard

Motor control is courtesy of the xESC project, which provides FOC motor control for low cost, interfacing with the host controller via a serial link. This is worth looking into in its own right! On the software side of things, [Clemens] is using ROS, which implements the low level robot control, path planning (using code taken from Slic3r) as well a kinematics constraints for object avoidance. The video below, shows how simple the machine is to operate — just drive it around the perimeter of lawn with a handheld controller, and show it where obstacles such as trees are, and then set it going. The mower is even capable of mowing multiple lawns, making the journey between them automatically!

Robotic mower projects are not new around here, here’s the mysterious TK with an interesting take, another using RTK GPS for good (or possibly bad) and quite probably the jankiest one we’ve seen in a while, which uses a LoRa base-station to transmit RTK corrections. We’d recommend keeping well away from that last one.

Continue reading “OpenMower: Open Source Robotic Lawn Mower With RTK GPS”

Draw On Your Lawn With This Autonomous Mower And RTK-GPS

The rise of open source hardware has seen a wide variety of laborious tasks become successfully automated, saving us humans a great deal of hassle.  Suffice to say, some chores are easier to automate than others. Take the classic case of a harmless autonomous vacuum cleaner that can be pretty dumb, bumping around the place to detect the perimeter as it traverses the room blindly with a pre-programmed sweeping pattern.

Now in principle, this idea could be extended to mowing your lawn. But would you really want a high speed rotating blade running rampant as it aimlessly ventures outside the perimeter of your lawn? The Sunray update to the Ardumower autonomous lawn mower project has solved this problem without invoking the need to lay down an actual perimeter wire. As standard consumer grade GPS is simply not accurate enough, so the solution involves implementing your very own RTK-GPS hardware and an accompanying base station, introducing centimeter-level accuracy to your mowing jobs.

RTK-GPS, also known as Carrier Phase Enhanced GPS, improves the accuracy of standard GPS by measuring the error in the signal using a reference receiver whose position is known accurately. This information is then relayed to the Ardumower board over a radio link, so that it could tweak its position accordingly. Do you need the ability to carve emojis into your lawn? No. But you could have it anyway. If that’s not enough to kick off the autonomous lawnmower revolution, we don’t know what is.

Continue reading “Draw On Your Lawn With This Autonomous Mower And RTK-GPS”

DIY Autonomous Mower In The Wild

Mowing the lawn is one of those repetitive tasks most of us really wish we had a robot for. [Kenny Trussell] mowing needs are a bit more strenuous than most backyards, so he hacked a ride-on mower to handle multi-acre fields all on it’s own.

The mower started out life as a standard zero turn ride on lawn mower. It’s brains consist of a PixHawk board running Ardurover, an Ardupilot derivative for ground vehicles. Navigation is provided by a RTK GPS module that gets error corrections from a fixed base station via an Adafruit LoRa feather board, to achieve centimetre level accuracy. To control the mower, [Kenny] replaced the pneumatic shocks that centred the control levers with linear actuators.

So far [Kenny] has been using the mower to cut large 5-18 acre fields, which would be a very time-consuming job for a human operator. A relay was added to the existing safety circuit that only allows the mower to function when there is weight on the seat. This relay is wired directly to the RC receiver and is controlled from the hand-held RC transmitter. It will also stop the mower if it loses signal to the transmitter. To set up mowing missions, [Kenny] uses the Ardupilot Mission Planner for which he wrote a custom command line utility to create a concentric route for the mower to follow to completely cover a defined area. He has made a whole series of videos on the process, which is very handy for anyone wanting to do the same. We’re looking forward to a new video with all the latest updates.

This mower has been going strong for two years, but in terms of hours logged it’s got nothing on this veteran robotic mower that’s been at it for more than two decades and still runs off an Intel 386 processor.

Where Are The Autonomous Lawnmowers?

It’s impossible to know when society began to manicure its front lawns. Truth be told — cutting the grass was, and still is a necessity. But keeping the weeds at bay, trimming, edging and so forth is not. Having a nice lawn has become a status symbol of modern suburbia all across the globe. When the aliens arrive, one of the first things they will surely notice is how nice our front lawns are. This feature of our civilization could have only been made possible with the advent of specialized grass-cutting machines.

reel mower
Reel Mower [Public Domain]
It could be argued that the very first lawnmowers were live stock. The problem was they were quite high maintenance devices and tended to provide a very uneven cut, which did not bode well for families striving for the nicest front lawn on the dirt road. Coupled with the foul odor of their byproducts, the animals became quite unpopular and were gradually moved out of site into the back yards. Other solutions were sought to maintain the prestigious front yard.

The first mechanical lawnmower was invented in 1830 by a man named Edwin Budding, no doubt in an effort to one-up his neighbor, who still employed a Scythe. Budding’s mower looked much like today’s classic reel mowers, where a rotating cylinder houses the blades and rotates as the mower is pushed forward. Budding was granted a patent for his device by England, much to the dismay of his fellow neighbors — most of whom were forced to buy Budding’s mower due to the fact that everyone else in the neighborhood bought one, even though they weren’t actually needed.

By the early 1930’s, the cold war started by Budding and his neighbor had spread to almost every front yard on earth, with no end in sight. Fast forward to the modern era and the lawn and garden market did 10 billion in sales in 2014 alone. Technological advances have given rise to highly advanced grass-munching machines. For smaller yards, most use push mowers powered by a single cylinder IC engine. Many come with cloth bags to collect the clippings, even though everyone secretly hates using them because they gradually fill and make the mower heavier and therefore more difficult to push. But our neighbors use them, so we have to too.  Larger yards require expensive riding mowers, many of which boast hydrostatic transmissions, which owners eagerly brag about at neighborhood get-togethers, even though they haven’t the slightest clue of what it actually is.

Us hackers are no different. We have front lawns just like everyone else. But unlike everyone else (including our neighbors) we have soldering irons. And we know how to use them. I propose we take a shot-across-the-bow and disrupt the neighborhood lawn war the same way Budding did 85 years ago. So break out your favorite microcontroller and let’s get to work!

Continue reading “Where Are The Autonomous Lawnmowers?”

Castrol Virtual Drift: Hacking Code At 80MPH With A Driver In A VR Helmet

Driving a brand new 670 horsepower Roucsh stage 3 Mustang while wearing virtual reality goggles. Sounds nuts right? That’s exactly what Castrol Oil’s advertising agency came up with though. They didn’t want to just make a commercial though – they wanted to do the real thing. Enter [Adam and Glenn], the engineers who were tasked with getting data from the car into a high end gaming PC. The computer was running a custom simulation under the Unreal Engine. El Toro field provided a vast expanse of empty tarmac to drive the car without worry of hitting any real world obstacles.

The Oculus Rift was never designed to be operated inside a moving vehicle, so it presented a unique challenge for [Adam and Glenn]. Every time the car turned or spun, the Oculus’ on-board Inertial Measurement Unit (IMU) would think driver [Matt Powers] was turning his head. At one point [Matt] was trying to drive while the game engine had him sitting in the passenger seat turned sideways. The solution was to install a 9 degree of freedom IMU in the car, then subtract the movements of that IMU from the one in the Rift.

GPS data came from a Real Time Kinematic (RTK) GPS unit. Unfortunately, the GPS had a 5Hz update rate – not nearly fast enough for a car moving close to 100 MPH. The GPS was relegated to aligning the virtual and real worlds at the start of the simulation. The rest of the data came from the IMUs and the car’s own CAN bus. [Adam and Glenn] used an Arduino with a Microchip mcp2515 can bus interface  to read values such as steering angle, throttle position, brake pressure, and wheel spin. The data was then passed on to the Unreal engine. The Arduino code is up on Github, though the team had to sanitize some of Ford’s proprietary CAN message data to avoid a lawsuit. It’s worth noting that [Adam and Glenn] didn’t have any support from Ford on this, they just sniffed the CAN network to determine each message ID.

The final video has the Hollywood treatment. “In game” footage has been replaced with pre-rendered sequences, which look so good we’d think the whole thing was fake, that is if we didn’t know better.

Click past the break for the final commercial and some behind the scenes footage.

Continue reading “Castrol Virtual Drift: Hacking Code At 80MPH With A Driver In A VR Helmet”