I Went Crazy During COVID Turning A Cheap R/C Car Into A Complex Solar-Powered Rover That Conquered The Australian Outback

For much of my life, I’ve always held a strong interest in robots, particularly those that we send out to explore. The idea that these machines can give us a window into a wider world always captivated me. Eventually, as my engineering skills and bank balance grew, I decided to finally pursue a passion project I’d been dreaming of since childhood. I was going to build a solar rover to send into the Australian outback, controlling it from hundreds of miles away in the comfort of my home.  It ended up being the most challenging project I’d ever taken on.

The idea for the project started all the way back in the 1990s. I was small child already obsessed with cars and electronics. One day, a family member showed me their latest creation—an RC car that could be driven via keypresses on an Amiga 500 computer. There was just one problem, eight-year-old me identified. You couldn’t drive it out of the room, because you couldn’t see where it was going. I figured a camera mounted on the car that fed back to the computer would be the perfect solution. The only problem was that technology wasn’t really there yet.

Fast forward to 2016. I’d recently moved to Melbourne to pursue an engineering career at a major automaker, and that saw me sorting through boxes full of personal items from years ago. Inside one was my Radio Shack Dodge Raminator RC car, complete with working suspension and a four-wheel-drive drivetrain. Seeing it again sparked memories in my head, and suddenly I realised that the time was ripe to build something great.

By this time, cellular data networks had spread across Australia. You could get decent bandwidth across most populated areas of the country on your mobile phone or with a USB modem. Webcams were cheap, and the Raspberry Pi Foundation had just launched the Raspberry Pi Zero, a computer the size of a stick of gum that could run on a mere trickle of electricity. Combine all these, and I realized you could make a remote control car that you could drive remotely from anywhere with mobile phone coverage. Throw a solar panel on top, and you could explore indefinitely. It sounded so easy and so compelling, I just had to try it.

The first steps weren’t even that hard! I bought myself a Raspberry Pi, along with some new batteries for the Radio Shack RC car. Unfortunately, it wasn’t built like a proper hobbyist-grade RC car, so interfacing it with the Raspberry Pi proved a little difficult. With some probing and investigation, I set it up so the Pi could trigger the drive motor in forward and reverse using the original RC car’s electronics and some transistors. This came with the limitation that there was no fine throttle control; it was either full forward, or full reverse. Radio Shack had also used some weird geared motor for the steering, which I ripped out entirely. I mechanically bodged in a standard hobby servo to the car’s steering linkage instead, which could be easily controlled by the Raspberry Pi.

On the recommendation of one of my best friends, an engineer named Alex, I chose to write the robot’s code in Python. Alex isn’t a software engineer or even particularly involved with computers, but he was right on the money. I was able to whip up code in a few days that let me drive the RC car over my home’s wireless network, giving me a glimmer of hope that this project could really work. Going by the name TK on YouTube, I nicknamed the little rover TKIRV—short for TK Internet Remote Vehicle.

The next step was getting this thing on the cellular network so I could drive it beyond the bounds of my home. I grabbed a cheap 3G USB dongle, and set up a VPN server at my house. This was necessary as the vast majority of consumer grade mobile data connections will not accept incoming connections. Thus, I couldn’t just look up the robot’s IP address and dial into it from my house. Instead, the robot would run a VPN client, and connect to the VPN server running on my home computer. This would then let me command the robot as if it was on a local network.

Getting a camera feed up proved more challenging. On my austere automaker salary, cash was tight, and I had to scrounge for parts. A local pawn shop had an old Microsoft webcam for $5 that ended up doing the trick. Hours of research and fussing around at the command line later, and I got video streaming up and running with an arcane tool called gstreamer.

Finally, all the pieces were in place. I had an RC car outfitted for remote control over a cellular data link and it had a camera on top. I dumped it in the driveway with Alex tailing along behind for safety, and sat myself down in my bedroom-cum-command center. Alas, our first attempts were to end in failure. The video feed kept dropping out, control was jerky and laggy, and we couldn’t go more than 30 feet without crashing. Several seconds would go by between me hitting the keys and the robot responding. It just wasn’t happening.

After hours of trying and failing, all seemed bleak, until a brainwave occurred. What if we just needed more bandwidth? We ran to the shops just minutes before close, and handed over a wad of cash for a faster USB modem. Some tinkering later, and we had the robot lashed up with its new faster data connection. I logged in with baited breath. Alex stood ready on the other end of the walkie-talkie, guarding the robot by the footpath.

“Oh my god, it works!” I exclaimed over the walkie-talkie as TKIRV took its first staggery steps down the road. With a few hundred milliseconds of video delay, and on-off throttle control, it wasn’t easy to keep it on the footpath. Regardless, this early test convinced me that the concept could work.

The overwhelming grind of full-time employment robbed me of time and joy over the years, but slowly and surely, I chipped away at the build. By early 2020, I’d upgraded the robot with finer throttle control with a hobby-grade ESC so it could drive more slowly instead of always zooming off into the weeds. I’d added a better camera and IR LEDs to give it a rudimentary ability to see in the dark. Test missions around the local neighborhood revealed further flaws, like a limited range of vision and vibrations ruining the view through the camera. More tweaks to frame rates and the streaming  software helped iron these out, along with a gimbal-stabilized camera mount that let the robot look left and right as well as straight ahead. Throwing a GPS on the robot helped me keep track of where it was and where it was going; I was able to pinpoint its location on Google Maps from back at base. With all this done, I had to undertake a major rebuild to implement the solar charging capability. With so many changes due to be made, I dubbed the new iteration SOLARBOI.

At this stage, I rebuilt the whole robot to properly accommodate a solar power system to recharge the batteries. This would give the robot the ability to stay out on a mission for days at a time. The main Raspberry Pi computer remained, responsible for talking to home base over the Internet, commanding the motor and steering, and handling the camera. A secondary Arduino microcontroller was then installed for power management reasons. The problem is that the power available from the solar panels wasn’t enough to keep the robot’s computer switched on and online at all times. Instead, the main Raspberry Pi would be powered down most of the time. The Arduino, which used far less power, would run a schedule, waking up the Raspberry Pi every few hours. The Raspberry Pi would connect to the internet over 4G, wait to see if I logged in, and go to sleep after 5 minutes if I didn’t. Keeping the main computer powered down most of the time massively reduced the robot’s daily power use, to the point where it could readily gain a good amount of charge from the sun each day.

This upgrade also saw the drivetrain and communications hardware split to run on separate batteries so that excess power draw from the drive motor wouldn’t compromise our 4G link back to base. The new design saw all the electronics sealed to prevent water ingress from rain and condensation, and the body got a coat of drab green paint for camouflage. Most of the upgrades were achieved with bits and pieces from my electronics junk box. The comms battery was a pack reassembled from a 1990s Toshiba laptop and some electrical tape. Camera brackets were fabbed up by hand out of old aluminium strip, because I had no 3D printer to do anything fancier.

I’d gotten to the point where it was time to leave the kiddie pool. No more tests in the driveway and the backyard. We had to go out bush and really put this thing to the test.

I picked out a route between two distant country towns as the first mission for my robot adventurer. The plan was to kick off in the first town, wind through the streets in the dead of night, and park on the outskirts to catch a charge from the sun during the day. Then, we’d start rolling again the next night on a hiking trail between the two towns, aiming to hit a distant phone box as our end goal approximately 10 miles away.

The idea excited and terrified me in equal measure. On the one hand, I was getting to fulfill one of my longest dreams, to send a robot out into the wild, entirely on its own, while I controlled it from a great distance away over the Internet. On the other hand, there was little stopping anyone from finding the robot out on its adventures and simply picking it up and stealing it away. At this point, I’d invested hundreds of hours building it and a not-insignificant sum of money. Still, that was the whole thing that made this enterprise so exciting—the sheer thrill of not knowing what would happen.

From the outset, I wanted these missions to be an event. I rigged up a Mission Control set at my house, and based my outfit on the NASA guys from Apollo 13. I spent the day of deployment driving a great distance to drop the robot off in the empty lot of a friendly local, then driving all the way home to get ready for the mission, to kick off in the dead of night. A first attempt faltered due to poor mobile signal in the distant country town, so I scrubbed the launch, did a ton more driving, and fitted upgraded antennas to the 4G modem. Finally, we were ready.

Friends tuned in to watch my robot’s first unaccompanied steps into the big wide world. Our first night out, we saw a kitty, struggled with the terrain, and got tangled in a wire fence. The second time, we got high-centered on a big rock stuck in the walking path. The poor field of view from the camera was making it difficult to drive, as was the dim output from my IR headlamps, and video lag of around half a second meant that progress had to be slow to avoid crashing into obstacles. Fog on the lens needed the fitment of a heater so we could still drive when the conditions got frosty.

Regardless, every mission was teaching me something, and I was more determined than ever. I was desperate to get the first multi-day mission in the bag, where we actually used the power of the sun to juice up the batteries for the next day’s attempt.

As COVID raged, Melbourne plunged into lockdown. Suddenly, effectively trapped in their homes, friends of friends were pouring into the Twitch streams to spectate and a little community formed to cheer SOLARBOI on. The missions weren’t always the most compelling to watch after the fact, but it was magic for those that tuned in live on Twitch. The idea of not knowing what was going to happen next, the drama, the suspense… we were seeing just what our little robot friend could do as he took his first steps into the world.

In mere weeks, I’d hit over 100 subscribers on the livestreaming platform, with over 20 concurrent viewers chatting away with advice, commentary, and speculation about what would happen next. A few weeks later, we blew past 30. It was a huge achievement for me just a month into streaming for the first time.

People dived in with offers of help. Popular chiptune musician and video expert ctrix helped rework my streaming setup. Pro coder Derwent X stepped up to upgrade my code, letting me hotkey all kinds of camera controls and other commands rather then having to waste time in the command line on stream. Telecoms expert Brett consulted on antenna upgrades and how to avoid noise impacting communications.

Donations poured in to help fund the project as I scrambled with every spare waking hour to improve the ‘bot. They came at a crucial time. COVID was raging, and business across Australia was feeling the pinch. Shortly after the first few missions, I was made redundant from one of my jobs, where I was designing a cutting-edge arcade machine. The business faced difficulty and with my first project complete, I was shown the door.  I had been on the verge of buying a house, and my life was suddenly thrown into shambles. I threw myself into my work on the robot, redoubling my efforts to make it all work. It was a troubling time, but SOLARBOI kept my hands busy and my mind occupied. Ultimately, I would end up overcommitting, but I had to see this through.

By our fourth time out, I was weeks into this crazy ride. Life was now about SOLARBOI, and SOLARBOI was life. The mission went well. We made it a good 2 kilometers the first night without too much trouble. The engineering updates to the bot had proven fruitful, our new high camera mount and fisheye lens was giving us a great view, and the wind was at our back. I parked up the robot out of the way, hoped nobody from the stream had figured out where it actually was, and went to bed.

I was so excited by the clean run, I could hardly sleep. The mission had lasted hours and I was spent but also peaking off caffeine. I’d been mainlining beverages from a Pyrex jug, which kind of became an in-joke in the streams. The next day I was too preoccupied to work. I kept frantically checking my back-end spreadsheets to which the robot was writing hourly location updates, to make sure nobody had stolen it. Battery voltages were slowly ticking up, showing the solar panel was doing its job. Eventually, in the early afternoon, I booted up a Twitch stream to see if we could really pull it off. What was, to the best of my knowledge at the time, the first DIY solar rover to undertake a multi-day mission under the power of the sun.

I logged in to the robot, raring to go. We had signal. The video feed came online. The viewers tuned in. I tapped out the commands, and it all came to fruition. SOLARBOI was a success. I cranked the tunes and screamed for a full minute at the glory of it all.

Of course, we still had a mission to do. Our objective lay a great distance away. Mission 5 saw SOLARBOI stretch a few hundred meters closer to its target, before I elected to stop to pick up more charge. Mission 6 saw us stretch ever further, a third day into proceedings. All was going well, until disaster struck. We were now a long way from the nearest mobile phone tower, with our tiny little antennas just a foot off the ground. When our proud robot rolled down an embankment, the steep roadside began to block our signal. I spent the best part of an hour trying to climb back up the steep slope, fighting a jagged video feed and minimal bandwidth. In a small boon, the camera heater did a great job of keeping the lens free of fog, but it wasn’t enough. I elected to park for the night, hoping another day’s charge might free our robot. Alas, it wasn’t to be. After driving our 20-year-old toy RC car 3.62 kilometers, it had dug itself into the dirt and the motor had ripped the drive pinion to pieces.

A series of fresh attempts followed as I fought the robot’s feeble Radio Shack underpinnings. I tried time again to upgrade the motors for more torque, locked the rear differential for more traction, and had friends donate new gears and other parts to try and beef up the drivetrain. I tried to fix communications issues with more hardware, better antennas, and more hardware. I desperately wanted to start afresh with a new design on a better chassis, but I felt like I didn’t have time. The fanbase was crying out for more regular content, and rebuilding the robot from the ground up would take months. I elected to keep trying running repairs with mixed success.

The robot grew heavier with each upgrade, and drivetrain failures became common. Hilariously, though, sometimes this was actually great fun. A successful multi-day mission in a new location actually saw us make some new friends, as I called the local publican to see if they could find my robot out in their backlot and rescue it from a sandy trap. I guided them via phone live on stream, while they wandered out into the paddock with a torch, listening out for guitar music playing from SOLARBOI’s speaker. Myself and the audience laughed like drains when we saw the torch beam glitter into view as our rescuers came across our stricken robot. We’d sheared another pinion gear, and the mission ended there, but it was a right laugh.

SOLARBOI being rescued by the local publican on its last bush mission. Aussies are always ready to help a robot in need.

From there, I retired SOLARBOI from public missions. The whole endeavor had become a huge strain on my life, both financially and mentally. I was spending morning to afternoon engineering upgrades to the ‘bot, before handing it off to a delivery driver to take it out to the countryside. I’d scarf down a quick dinner, spend a couple hours tweaking code, and then start broadcasting at around 9 or 10PM. The late-night missions were propelled by caffeine and sheer elation, and then I was generally too tired to sleep before 5AM. I’d spend the next day obsessively checking the robot’s vitals online, or I’d be on a recovery mission to pick it up and fix whatever broke. It wore me ragged, and it was time to take a break.

It’s hard to focus on real life when you’re spending every spare moment working on a complicated project. It’s even worse when you’re spending all day stressing about your multi-thousand-dollar robot and whether some passing stranger has chosen to nick off with it. In any case, by this point, the toy car I’d built it on was beyond repair. After a few experiments letting Twitch drive the bot, I chose to turn SOLARBOI into MOWERBOI, repurposing it with an autonomous GPS guidance system that would see it mowing my back lawn for me. That project itself taught me a great deal about autonomy and navigation, and left me wanting more. I still have a stack of videos of how that project ended up that needs to see the light of day.

I’ve since sourced a far more capable off-road platform for a future SOLARBOI 2.0. I’d love to invest the time and money into building a new revision from scratch. It would have a better camera, better night vision, and the ability to guide itself on a GPS path to solve some of our issues with driving on video delay. Plus, it would be far lighter and lower to the ground for better handling and better efficiency. Most of all, unlike the original toy-based version, it would have a stout high-torque drivetrain built for crawling along and dealing with soft terrain. That alone would be the biggest upgrade over the initial build, which was meant for tearing around at speed on hard surfaces, not fighting through coarse Australian sand and mud.

Given the opportunity and the engineering time, I hope to one day see SOLARBOI roam farther and wider. I’d love to complete a week-long mission and see it break the 10-, or even 20-mile barrier. With the right location, ideally free of cars and nosy interlopers, I think I could achieve that and even more. Here’s to the future, and long may SOLARBOI roam.

Image credits: Lewin Day, What Up TK Here via YouTube screenshot