Start by deciding its purpose. The form will follow the purpose. If you know electrical engineering, you won't make as many mistakes as I have. If you know mechanical engineering, you won't take as long to design the components as it took me. If you're a competent programmer, you won't be as slow integrating the hardware into a cohesive whole as I am. If you've read The Design of Everyday Things, you're ahead of the curve in terms of making it accessible.
A good starting point is learning to use ROS https://ros.org/. Once you learn ROS you can interact with/use a wide variety of robots. Robotics is a fairly multidisciplinary field though, so it really depends on what area of robotics you are interested in.
I'm not joking, but some of those Lego robotics kits can work just fine for getting your feet wet. Yes, even if they are made for kids.
When I started out, it was with microcontrollers, breadboards, sensors, actuators. And even though it was in college, I do wish I had just played around with more high-level stuff first.
The lego mindstorms kits were great for this. Nice and modular for the hardware, but still real enough to teach you lessons like sometimes identical motors don't work the same way.
Although the Mindstorms line is discontinued, nowadays there's the Lego SPIKE Prime (it's marketed more towards classrooms). Like the Mindstorms EV3, you can program it with either a visual block language or MicroPython.
Robotics is more accessible than ever. The tech is here to build almost anything you want. Dream big! We can now buy cameras for a couple dollars, microphone arrays, sensors, motors and drivers. 3D printers are everywhere, stock components also available.
Or 'manage your expectations', download ROS and build another mediocre turtle bot powered by some Nvidia Teraflop chip (if they allow you to boot their sdk), and spend a year learning "abstractions" defined by other people for other projects.
Has anybody tried Crunch Lab's HackPack? It seems it could be quite well suited for this purpose -- they claim to make it straight forward to go beyond the kit, and they've certainly taken some of the right steps in that direction.
I tried it. First kit was speced with cheap parts that didn’t really work. Got a replacement kit - same thing. Overall a really disappointing experience. I cancelled and bought a 3D printer, which has been great fun.
Totally free, don't need to buy anything just a computer.
I've been designing electrical hardware for robots for the last 4 years for a big corporation and I can tell you, the fun and money is in the software but having another skill is awesome. Robotics is a place where multiple fields converge and if you find a good team they will help you to grow.
Embedded is the adjacent field after you complete the course. Maybe something like zephyr project.
If you want to get you feet wet with mechanical. Buy a A1 mini and play with onshape (www.onshape.com) to design your first pieces, supports for the motors or the board, try create you own gripper.
As for the electrical engineering, is the one with most pitfalls and the most expensive. A wrong voltage will release the magical smoke and is another 30$ for a board. Tread carefully. Start with the RP2040 or the RP2350, they are cheap and well documented. This skill will evolve hand by hand with the embedded coding. Start small. Learn about H-bridge and brushed motors before doing the jump to the bigboys and FOC control with brushless. Get a cheap soldering iron. If you can, a clone of the JBC C245 tips. Is the most versatile and you can find stuff in alie xpress for 45 or 50$ and would be similar to the tools you will find in the field without breaking the bank.
Search for ROS meetups. I could point you to some depending where in the world you are.
And above all, it will be a long journey. Don't dispear, do at your own time but don't forget the objective.
Microcontrollers and PLC are historically different fields that have some overlap as the fields converge somewhat. PLC come from relay logic and so learning Ladder Logic and playing with a cheaper Automation Direct PLC might be good if you're looking to do control systems like industrial machines and those robotics. RPi series micros are good to learn embedded electronics, servo controls and Realtime OS. Computer Engineering and Control System Engineering are treated differently even in the California State Engineering License Exams, so they tend to be different paths but the skills have some overlap. My degree was in computer engineering and didn't cover any PLC, but I contract and PLC comes up a lot in one-off things that need to be reliable.
Yeah. As some has said already PLC are more oriented to robotic arms in factories etc, where space is not a premium. Robots like anymal from anybotics, spot from boston dynamics or the unidog from unitree to name a few don't use PLC at all due to lack of space. Also energy efficiency, mcu at most will push less than a watt to move something complex, a motor control for the FOC, I mentioned earlier.
I worked for the past 4 years in HP siteprint.The main problem was developing all to fit inside a pelican case that you could check in on a plane. The smallest PLC was almost the full internal volume of the robot. That's why I said microcontrollers, more ready available to experiment at home, faaaaaaaar more cheaper and those two, very well documented. STM32 in europe is the big player but is harder to get into, the documentation tends to be cahotic.
Though by total coincidence, it looks like I'm about to inherit responsibility for a fleet of Raspberry Pies that manage dashboards on a factory production floor. But they are older, and it's probably just going to be like managing any other low spec Debian server, albeit connected to a monitor.
Mmmm true, but will not replace a good coder. The software team has been experimenting with AI. Is useful and can guide you to do a Proof of concept fast but if you want to start doing very specific custom stuff all crumbles down fast.
In the next years? Yeah I see resolving path planning without a doubt or reverse kinematics or object recognition, for example. AI will be another tool, one you have to learn to use it. But replacing a full software guy... I don't see it, to be honest. I see it more like in the novel Beyond the Blue Event Horizon. The first chapters explains AI as the AI agents that are being rolling out nowdays. But they will be more like what a junior engineer is to a senior, more than replacing a full team of engineers.
AI trivializes coding: the complexity and scope of robot software will increase.
Robotics-oriented ML trivializes current robot design problems: demand for robotics engineers increases, the projects do cooler things but the day-to-day gets less interesting.
All we really need for the robotics AI revolution is to deploy many tens of thousands of well built, cheap, reliable robots sold at a profit that customers get tons of utility out of, get petabytes of data back from them while complying with data privacy regs, train a model on this data on a few thousand H200s, and then deploy that model back to the robots, repeating the cycle every few weeks. The rest will be easy.
I'm in the field and I disagree with the initial direction of this post. You aren't going to 'learn robotics' in a meaningful manner by checking the boxes in a online ROS2 course.
Robotics is a compound discipline which pairs mechanical and electrical engineering knowledge with mathematics and software development. In order to get meaningful practice in all of these areas, I would recommend re-implementing a robot vacuum from 'scratch'.
The actual vacuum part is non-essential, as you're really trying to implement the famous 'turtle' robot and experience all the design steps/compromises/challenges first hand. As a goal, aim for autonomous navigation around your home that works equivalently to a cheap 'throwaway' robo-vac.
Keep in mind you are surrounded by robotics. Most passenger vehicles, consumer drones, micro-mobility devices and modern construction equipment are 'robots in varying degrees of disguise'. Practical knowledge regarding any of these systems transfers well if you know where to apply it.
Where would you recommend someone to start from a more data-sciencey background, algorithms like VSLAM look interesting to mess around with but building a whole robot just to do so seems daunting.
It's similar, but the advice should be good for a software professional. It's an actionable, well-documented goal which takes you through the basics you're likely to care about. If any step needs some extra explanations, the choice of project guarantees you can find hundreds of good resources for that step in particular. With that basic background, next steps will be easier to find and understand.
There was no advice provided other than to "look it up". That's not particularly helpful for someone new to the field and overwhelmed by the amount of resources and approaches. If you are aware of "documentation" it would be more useful to provide links to such resources for someone just getting started.
A solid resource for me getting started was the textbook Probabilistic Robotics. It may not be state of the art but I feel it is a great resource on the topic -
I did a very basic version of this and it was a lot of fun. I bought a chassis online with tank treads that each had a motor. I used an Arduino to send the signal to each tread to turn and a distance sensor so it knew when it was arriving at an object. If it sensed that something was in front of it, it just turned 90 degrees and kept going.
Super simple but it felt like a big accomplishment to get that far.
> As for the electrical engineering, is the one with most pitfalls and the most expensive. A wrong voltage will release the magical smoke and is another 30$ for a board. Tread carefully.
I enjoyed reading "Exploring Beaglebone" by Derek Molloy a few years ago, because the book has sections "how to destroy your beaglebone" and "how to interface with electronics". The website seems broken. The book's ISBN is 978-1-118-93512-5. Some of the recommendations there (voltage clamping circuit on the 1.8 V analog inputs) can save your board.
> And above all, it will be a long journey. Don't dispear, do at your own time but don't forget the objective.
My biggest issue with robotics is that I don't see the point in any objective.
Building a cool robot feels like it's just a toy to me, and I don't know how to get around that feeling. I studied mechatronics at uni and spent a long time learning on my own in addition.
Robotics for business is almost a cliff, very hard to get into and make something that works reliably at a reasonable cost, from a solo developer point of view
>very hard to get into and make something that works reliably at a reasonable cost, from a solo developer point of view
I mean yes, otherwise they'd be even more robotic stuff wandering about
In products they have the trade off triangle where you can pick things like fast-cheap-good, pick any two.
In robots it's cheap-reliable-featureful pick any one.
The problem with the term 'cool' is it sets no boundaries. Those constraints limit your costs and scope. That's the hard part for most people when they think about cool.
And reliable is more of a time factor (+effort). QA on products takes a massive amount of time and effort to figure out what is not reliable.
Robotics is definitely getting more accessible! But keep in mind it’s a whole different world from traditional web or desktop software development, so don’t be surprised by the relatively steep learning curve. Keep at it though!
- Buy a small robot kit from Amazon or a local reseller. Yahboom make some good robot toy car kits. Hugging Face have the open source SO-ARM101 that plenty of companies manufacture and sell now. Expect to spend about $250 USD including a Jetson Nano for a good kit, up to $1000 USD if you want some more sensors
- If you can’t afford a real robot, play around with simulators like Isaac Sim and Mujoco
- Check out LeRobot, excellent framework for ML robotics from Hugging Face
- Learn the basics of ROS
(pubsub), even if you don’t end up using it, a lot of the industry jargon and design patterns come from ROS so it helps to understand it. Think of ROS like Ruby on Rails, it’s a heavyweight batteries-included framework with lots of opinions.
- ROS does have some nice libraries for manipulation (MoveIt) and navigation (Nav2) using more classical (non-ML) methods
- Leverage AI tools such as ChatGPT and Cursor when you get stuck, it’s a lot faster than Googling when you’re just getting started and don’t even know the right term to search for.
- (Shameless plug) Check out two tools I’m working on: mcap.dev for logging and foxglove.dev for visualization
Maybe ROS is the way to get started in robotics and for AGVs it's probably the most resourced thing one could learn.
But that is really only for AGVs. Once you get into the air, it's all Mavlink (always has been), ASVs and AUVs still run MOOS, and robotic arms are far more likely to use ABB studio than Moveit and the research is going towards LeRobot with end-to-end control, not classical methods. IoT systems are mostly Home Assistant, and well, some just use MQTT directly instead.
ROS is probably not as ubiquitous as us ROS devs would like to believe, what one should learn really depends on the field one wants to be in.
ROS is still very heavily used in industry and is growing in absolute terms, but most of the newer frameworks and research are learning towards e2e methods.
But even if you don't use ROS, it really helps to have a working knowledge of how it approaches things like pubsub, logging + replay, transforms, etc.
I highly recommend starting with the SO-ARM101 and the LeRobot tutorial. They're super cheap, its insanely quick to get started, and you can even buy pre-made kits like at https://partabot.com . It's the "Hello World" of robotics now, imo.
Don't bother with a Jetson Nano, you don't need that to get started, and by the time you need that you'll know a lot already. You can just drive the robot from your laptop!
Getting to training your own VLA fine-tuned model is a super quick and easy process. You can see examples of other people completing the tutorial and uploading their training/evaluation datasets here (shameless plug for my thing): https://app.destroyrobots.com
I wouldn't bother much with ROS at first tbh. It'll bog you down, and startups are moving toward using other approaches that are more developer friendly, like Rust-based embedded.
You can go far with a robot connected to USB though!
There are many ways to start with a more academic knowledge of robotics. Courses, ROS, Programs etc. Just picking one is a fine start and you will learn something, but you may not have fun or appreciate the context for what you are learning, which makes it hard to care.
It also depends on what you're trying to learn. Programming? control? electronics? mechanism design? There are many different aspects that require their own expertise. I personally think of a roboticist as someone that can be functional in all the disciplines and bring it all together. Some one that can build a robot from scratch.
This advice is targeted towards someone starting in robotics with an emphasis on building a totally custom robot. You are doing the whole thing from scratch, no kit. (kits are super useful, but I personally like building everything and making hardware that works for me). In my eyes, if you haven't built it in hardware, it probably isn't a robot.
As a research roboticist, my best advice is forget Learning it all up front and just start building. You will probably have strengths and weaknesses, but robotics is a very multi-disciplinary problem and at the start, requires some head banging in what we like to call integration hell. Just Trying to build something does a couple things. It gets you acquainted with the interaction of all aspects (mechanical, electrical, software) of a robot and how they interact. Most importantly it gives you context for sll the things you learn later on. Build something, come across a problem, and google/brute force your way to a solution. You will stumble across context relevant resources and lessons in the moment. Especially at the start, a more "formal" education is too broad and comprehensive for the curious roboticist.
My simple advice is to just come up with something simple and SEND IT!
For mechanical - Onshape for 3D design (free tier) and a bambu A1 mini 3d printer can get you building anything you want for <$400. A couple nights of CAD tutorials will get you started. Hell you can even start with cardboard and hot glue if this is cost/time prohibitive.
For electronics - Browse www.adafruit.com or www.sparkfun.com for sensors and microcontroller boards. They both support QWICC connections which makes getting sensors plugged in and running super easy for a beginner. For computation and control a basic Arduino Uno or one of the adafruit feathers will do just fine. Don't get caught up in the internet debate on what the best microcontroller/single board computer is. The best robot is one that works and it is VERY easy to get in over your head in this realm and fall into the premature optimization track. Better to get something working that under performs than to never build at all.
For software - Most lower level robotics uses c++, which I think scares people. It is mostly because it is one level above hardware. Arduino/common libraries take care of all the hard stuff and you don’t have to worry about large scale things like garbage collection, memory leaks, or efficient code off the bat. The websites above have libraries, examples, and tutorials for every sensor. Even a non programmer type can smush the example code together in a functional way and get something up and running.
As you try to build you will learn and get more complicated. Robotics is so vast and large its best to focus on learning the skill you need at the time. But my main advice is just go build the thing and the learning resources/skills will present themselves when they are needed. Otherwise it is just too much to take a more planned approach.
The electronics learning is always going to apply to start there if you need to, then depending on your goals the hardware and software will vary.
For simple DIY projects I'd start with getting an Arduino or Arduino compatible board and just play around with building things with sensors, LEDs, motors, etc. You'll find lots of kits and resources online for simple Arduino projects.
More complex projects are probably going to need to be Linux-based (AI, machine vision, etc). In which case you might want to look at building something with a mini PC, Raspberry Pi, NVIDIA Jetson.
These days robotics is super accessible, but can cost a bit depending on what you want to do... That said, most of the time you'll find there are cheap ways to build POCs to validate your ideas first.
If you want to understand how it works, take first Stanford's CS223A and then MIT's 6.832 Underactuated Robotics courses. They have lectures online as well. Udacity's Self-driving Car Nanodegree might be a good idea after those two.
Everyone is jumping the gun here. You just haven't told us enough.
How much time do you have? How much do you know about hardware and software? What do you want to do? How much money do you want to invest? What do you want to learn?
The path forward changes entirely depending on the answers to these questions.
It's not that hard to get started! I make robots like this (that one's just to enter the wiring stage) using alu extrusion beams, some 3D printing, off-the-shelf servos and SBCs:
Modern SBCs are decently powerful and run for long enough off a standard drone LiPo pack that you can fairly easily build yourself a solid little platform to learn and experiment with different controller approaches. Stereo cams, LiDAR modules, etc. are also readily available at a quality level that is decent enough.
You'll quickly realize the amount of work needed to bring any of the more advanced approaches to a productization level, but a "I implemented enough of this to get the principle" mini-Boston-Dynamics is in reach of the hobbyist.
In fact, the programming craft aspect of all of this is relatively simple. Basic implementations are typically not architecturally complex or a lot of sloc. You'll find that it's more about learning some control theory, and it'll also get you into machine learning.
A really top-tier robotics engineer is going to be a generalist across mechanical, electrical, and several software domains, and then also a specialist, deep expert in one or two of those. So I'd recommend doing a bit of everything then picking a specialty to learn in depth. Build a really simple kit robot and then program it to do simple stuff using ROS. Write a few ROS nodes from scratch to do something simple, things that existing ROS nodes already do. Then learn why the STL ROS options are (most likely) better than yours.
I strongly recommend taking the generalist approach. It makes you a much better engineer in the long run. I had a friend call me up enraged one day. He got woken up at 1am and ordered to drive across the state to get a site back up and running because his hardware was broken and the very expensive software team was sitting around wasting their time. So he got to the site at three or four AM and found that the room full of software engineers, three PhDs among them, didn't understand that BOTH of the battery terminals needed to be connected. Yes, batteries in fact have a positive and a negative. And these people were each getting paid twice as much as him. He left that company and became hardware leadership at a much better company, thank goodness. My point is, super-specializing in the most profitable hyper-niche right off the bat is going to make you a very mediocre robotics engineer who might be successful in the short term, but might find yourself unemployed in a few years when we find a better modality to replace your hyper-niche. Specialists win the sprints, generalists win the marathons.
And if you're in high school, join a First Robotics team. (In First Robotics, teams build a moderately large robot that can zip around a field, grabbing things and shooting balls for instance. It's like a competitive sport but with robots competing. The robots are controlled by drivers, but can also act autonomously for more points. The task is different each year, with sub-tasks of varying levels of difficulty, so teams with widely different skill levels can take part: some robots are very simple while others are amazingly complicated. It's a great way for high schoolers to learn engineering skills, anything from programming to mechanical engineering.)
I would say for starting a first robotics lego league is going to be easier to access because they build a smaller robot out of legos, plus the "season" starts in the Fall vs. January for First robotics league.
I've coached a First team that has qualified to go to worlds the last four years and been on the Einstein field once in that time. I think you'd learn a few things
- robots are built to solve problems and the most technically complex and advanced robots don't always solve problem better than simple elegant designs.
- Robots involve multiple skills and usually get built by teams where members are specialists in specific fields. Some people will understand all of the robot sub systems but
FIRST and most of its teams are very open to collaborating and supporting rookies new to the org. I recommend reaching out to as many nearby teams as possible and see what support they can provide. Any serious teams are motivated to help as this directly contributes to their team's success during the competitive season. Competitions require a teams to have a pro-social aspect to their operations. Helping other teams is a strong signal in this area.
Additionally, find a nearby FRC competition and volunteer for at least one event (do more, if you can). Wander the robot pit and interact with the teams. There will be a lot of good intel for you there just wandering around and asking questions.
I couldn't agree more. FIRST is an excellent organization to get involved with. Even if you don't have specialities which directly align with the needs of a team, most will not turn you away (especially if you demonstrate passion).
Just a few thoughts about starting a team and/or volunteering:
Starting a FIRST Robotics Competition (FRC) team is not for the feint-of-heart and requires significant time and money investment. This is the sort of endeavor you try when you're either retired or have a group of other dedicated volunteers willing to help you build the team over multiple seasons. If you spend a year or two mentoring/volunteering for a well-established team, you'll get a good sense of what you're getting into. FRC is a bit hardcore.
If no FRC teams exist nearby, FIRST Lego League (FLL) is a good entrypoint, but may not be technical enough for an adult interested in using FIRST to gain exposure to robotics. (It is plenty satisfying to mentor these teams, but Mindstorms can be somewhat limiting.) These are geared toward using Lego Mindstorms for learning robot concepts.
A good middle ground (for exposure to more practical robotics) would be volunteering with a FIRST Tech Challenge (FTC) team. It's the middle ground between FLL and FRC and is _doable_ for a single, dedicated, passionate adult to start (though I'd strongly recommend finding another volunteer to pursue this with).
If getting involved with a team is not an option, seek out nearby FIRST events that you can volunteer at. Many competition events need volunteers and you'll get a chance to interact with local teams and find plenty of opportunities to play with robots.
As someone in the industry (I literally can call myself a Robot Master having a Masters Degree in Robotics), these simple “kid kits” are phenomenal. When you first start out, the hardest part is knowing the hardware is actually going to do what you tell it to do. Simplifying the mechanical engineering/electrical engineering so you can immediately create real world actions makes a huge difference in giving you motivation to keep going.
Once you have the basics in place and you understand mechanisms, sensors, and some algorithms, you can move on to other parts of robotics.
However, keep in mind robotics is really about system thinking… you have to understand trades between sensing, thinking, and acting. But most importantly you have to remember you are dealing with the real world… you can’t as easily write a unit test and know that a particular actuator will behave exactly the same way every time.
The best way to learn is just by running small experiments and being curious about the field.
Just a few weeks ago, along with my partner i started this Sunday experiment of building and learning robotics as a software engineer.
We ordered a one hand robotic arm with orin nvidia gpu to start running experiments.
to commit and make it a habit, we also started documenting our learnings and experience.. if this motivates you you can check it out here - https://youtu.be/pAwkl6zd0ws
105 comments
[ 3.9 ms ] story [ 188 ms ] threadThere will be a world wide hackathon in two weeks time, no better way to get started and get to know some people: https://huggingface.co/LeRobot-worldwide-hackathon
When I started out, it was with microcontrollers, breadboards, sensors, actuators. And even though it was in college, I do wish I had just played around with more high-level stuff first.
IMO https://pybricks.com/project/saving-lego-mindstorms/ is especially interesting.
Or 'manage your expectations', download ROS and build another mediocre turtle bot powered by some Nvidia Teraflop chip (if they allow you to boot their sdk), and spend a year learning "abstractions" defined by other people for other projects.
Edit - fixed autocorrect
Totally free, don't need to buy anything just a computer.
I've been designing electrical hardware for robots for the last 4 years for a big corporation and I can tell you, the fun and money is in the software but having another skill is awesome. Robotics is a place where multiple fields converge and if you find a good team they will help you to grow.
Embedded is the adjacent field after you complete the course. Maybe something like zephyr project.
If you want to get you feet wet with mechanical. Buy a A1 mini and play with onshape (www.onshape.com) to design your first pieces, supports for the motors or the board, try create you own gripper.
As for the electrical engineering, is the one with most pitfalls and the most expensive. A wrong voltage will release the magical smoke and is another 30$ for a board. Tread carefully. Start with the RP2040 or the RP2350, they are cheap and well documented. This skill will evolve hand by hand with the embedded coding. Start small. Learn about H-bridge and brushed motors before doing the jump to the bigboys and FOC control with brushless. Get a cheap soldering iron. If you can, a clone of the JBC C245 tips. Is the most versatile and you can find stuff in alie xpress for 45 or 50$ and would be similar to the tools you will find in the field without breaking the bank.
Search for ROS meetups. I could point you to some depending where in the world you are.
And above all, it will be a long journey. Don't dispear, do at your own time but don't forget the objective.
I worked for the past 4 years in HP siteprint.The main problem was developing all to fit inside a pelican case that you could check in on a plane. The smallest PLC was almost the full internal volume of the robot. That's why I said microcontrollers, more ready available to experiment at home, faaaaaaaar more cheaper and those two, very well documented. STM32 in europe is the big player but is harder to get into, the documentation tends to be cahotic.
Thank you!
Though by total coincidence, it looks like I'm about to inherit responsibility for a fleet of Raspberry Pies that manage dashboards on a factory production floor. But they are older, and it's probably just going to be like managing any other low spec Debian server, albeit connected to a monitor.
This could change if there are breakthroughs in AI, which is not unlikely in the coming few years. Any thoughts on that?
In the next years? Yeah I see resolving path planning without a doubt or reverse kinematics or object recognition, for example. AI will be another tool, one you have to learn to use it. But replacing a full software guy... I don't see it, to be honest. I see it more like in the novel Beyond the Blue Event Horizon. The first chapters explains AI as the AI agents that are being rolling out nowdays. But they will be more like what a junior engineer is to a senior, more than replacing a full team of engineers.
Robotics-oriented ML trivializes current robot design problems: demand for robotics engineers increases, the projects do cooler things but the day-to-day gets less interesting.
Both: this is probably just AGI right?
Robotics is a compound discipline which pairs mechanical and electrical engineering knowledge with mathematics and software development. In order to get meaningful practice in all of these areas, I would recommend re-implementing a robot vacuum from 'scratch'.
The actual vacuum part is non-essential, as you're really trying to implement the famous 'turtle' robot and experience all the design steps/compromises/challenges first hand. As a goal, aim for autonomous navigation around your home that works equivalently to a cheap 'throwaway' robo-vac.
Keep in mind you are surrounded by robotics. Most passenger vehicles, consumer drones, micro-mobility devices and modern construction equipment are 'robots in varying degrees of disguise'. Practical knowledge regarding any of these systems transfers well if you know where to apply it.
It's like saying how do I learn car manufacturing but really I just want to try to make rubber tyres.
A solid resource for me getting started was the textbook Probabilistic Robotics. It may not be state of the art but I feel it is a great resource on the topic -
[1] https://robots.stanford.edu/probabilistic-robotics/
This is the best answer-- all other answers are saying "Run this simulation of a robot".
but a simulation is not a robot. Even a simple lego tupe robot kit shows you how fiddly and complex real world moving parts make everything.
Super simple but it felt like a big accomplishment to get that far.
I enjoyed reading "Exploring Beaglebone" by Derek Molloy a few years ago, because the book has sections "how to destroy your beaglebone" and "how to interface with electronics". The website seems broken. The book's ISBN is 978-1-118-93512-5. Some of the recommendations there (voltage clamping circuit on the 1.8 V analog inputs) can save your board.
My biggest issue with robotics is that I don't see the point in any objective.
Building a cool robot feels like it's just a toy to me, and I don't know how to get around that feeling. I studied mechatronics at uni and spent a long time learning on my own in addition.
Robotics for business is almost a cliff, very hard to get into and make something that works reliably at a reasonable cost, from a solo developer point of view
I mean yes, otherwise they'd be even more robotic stuff wandering about
In products they have the trade off triangle where you can pick things like fast-cheap-good, pick any two.
In robots it's cheap-reliable-featureful pick any one.
The problem with the term 'cool' is it sets no boundaries. Those constraints limit your costs and scope. That's the hard part for most people when they think about cool.
And reliable is more of a time factor (+effort). QA on products takes a massive amount of time and effort to figure out what is not reliable.
Noted not just for pedantry but you may get more attention.
This is an example: https://youtu.be/mTpkV7xZln0
- Buy a small robot kit from Amazon or a local reseller. Yahboom make some good robot toy car kits. Hugging Face have the open source SO-ARM101 that plenty of companies manufacture and sell now. Expect to spend about $250 USD including a Jetson Nano for a good kit, up to $1000 USD if you want some more sensors
- If you can’t afford a real robot, play around with simulators like Isaac Sim and Mujoco
- Check out LeRobot, excellent framework for ML robotics from Hugging Face
- Learn the basics of ROS (pubsub), even if you don’t end up using it, a lot of the industry jargon and design patterns come from ROS so it helps to understand it. Think of ROS like Ruby on Rails, it’s a heavyweight batteries-included framework with lots of opinions.
- ROS does have some nice libraries for manipulation (MoveIt) and navigation (Nav2) using more classical (non-ML) methods
- Leverage AI tools such as ChatGPT and Cursor when you get stuck, it’s a lot faster than Googling when you’re just getting started and don’t even know the right term to search for.
- (Shameless plug) Check out two tools I’m working on: mcap.dev for logging and foxglove.dev for visualization
But that is really only for AGVs. Once you get into the air, it's all Mavlink (always has been), ASVs and AUVs still run MOOS, and robotic arms are far more likely to use ABB studio than Moveit and the research is going towards LeRobot with end-to-end control, not classical methods. IoT systems are mostly Home Assistant, and well, some just use MQTT directly instead.
ROS is probably not as ubiquitous as us ROS devs would like to believe, what one should learn really depends on the field one wants to be in.
ROS is still very heavily used in industry and is growing in absolute terms, but most of the newer frameworks and research are learning towards e2e methods.
But even if you don't use ROS, it really helps to have a working knowledge of how it approaches things like pubsub, logging + replay, transforms, etc.
Don't bother with a Jetson Nano, you don't need that to get started, and by the time you need that you'll know a lot already. You can just drive the robot from your laptop!
Getting to training your own VLA fine-tuned model is a super quick and easy process. You can see examples of other people completing the tutorial and uploading their training/evaluation datasets here (shameless plug for my thing): https://app.destroyrobots.com
I wouldn't bother much with ROS at first tbh. It'll bog you down, and startups are moving toward using other approaches that are more developer friendly, like Rust-based embedded.
You can go far with a robot connected to USB though!
It also depends on what you're trying to learn. Programming? control? electronics? mechanism design? There are many different aspects that require their own expertise. I personally think of a roboticist as someone that can be functional in all the disciplines and bring it all together. Some one that can build a robot from scratch.
This advice is targeted towards someone starting in robotics with an emphasis on building a totally custom robot. You are doing the whole thing from scratch, no kit. (kits are super useful, but I personally like building everything and making hardware that works for me). In my eyes, if you haven't built it in hardware, it probably isn't a robot.
As a research roboticist, my best advice is forget Learning it all up front and just start building. You will probably have strengths and weaknesses, but robotics is a very multi-disciplinary problem and at the start, requires some head banging in what we like to call integration hell. Just Trying to build something does a couple things. It gets you acquainted with the interaction of all aspects (mechanical, electrical, software) of a robot and how they interact. Most importantly it gives you context for sll the things you learn later on. Build something, come across a problem, and google/brute force your way to a solution. You will stumble across context relevant resources and lessons in the moment. Especially at the start, a more "formal" education is too broad and comprehensive for the curious roboticist.
My simple advice is to just come up with something simple and SEND IT!
For mechanical - Onshape for 3D design (free tier) and a bambu A1 mini 3d printer can get you building anything you want for <$400. A couple nights of CAD tutorials will get you started. Hell you can even start with cardboard and hot glue if this is cost/time prohibitive.
For electronics - Browse www.adafruit.com or www.sparkfun.com for sensors and microcontroller boards. They both support QWICC connections which makes getting sensors plugged in and running super easy for a beginner. For computation and control a basic Arduino Uno or one of the adafruit feathers will do just fine. Don't get caught up in the internet debate on what the best microcontroller/single board computer is. The best robot is one that works and it is VERY easy to get in over your head in this realm and fall into the premature optimization track. Better to get something working that under performs than to never build at all.
For software - Most lower level robotics uses c++, which I think scares people. It is mostly because it is one level above hardware. Arduino/common libraries take care of all the hard stuff and you don’t have to worry about large scale things like garbage collection, memory leaks, or efficient code off the bat. The websites above have libraries, examples, and tutorials for every sensor. Even a non programmer type can smush the example code together in a functional way and get something up and running.
As you try to build you will learn and get more complicated. Robotics is so vast and large its best to focus on learning the skill you need at the time. But my main advice is just go build the thing and the learning resources/skills will present themselves when they are needed. Otherwise it is just too much to take a more planned approach.
- Electronics (batteries, resistors, capacitors, etc) - Robotic hardware (motors, sensors, microcontrollers, etc) - Software
The electronics learning is always going to apply to start there if you need to, then depending on your goals the hardware and software will vary.
For simple DIY projects I'd start with getting an Arduino or Arduino compatible board and just play around with building things with sensors, LEDs, motors, etc. You'll find lots of kits and resources online for simple Arduino projects.
More complex projects are probably going to need to be Linux-based (AI, machine vision, etc). In which case you might want to look at building something with a mini PC, Raspberry Pi, NVIDIA Jetson.
These days robotics is super accessible, but can cost a bit depending on what you want to do... That said, most of the time you'll find there are cheap ways to build POCs to validate your ideas first.
How much time do you have? How much do you know about hardware and software? What do you want to do? How much money do you want to invest? What do you want to learn?
The path forward changes entirely depending on the answers to these questions.
https://i.imgur.com/9YIHT9P.jpeg
Modern SBCs are decently powerful and run for long enough off a standard drone LiPo pack that you can fairly easily build yourself a solid little platform to learn and experiment with different controller approaches. Stereo cams, LiDAR modules, etc. are also readily available at a quality level that is decent enough.
You'll quickly realize the amount of work needed to bring any of the more advanced approaches to a productization level, but a "I implemented enough of this to get the principle" mini-Boston-Dynamics is in reach of the hobbyist.
In fact, the programming craft aspect of all of this is relatively simple. Basic implementations are typically not architecturally complex or a lot of sloc. You'll find that it's more about learning some control theory, and it'll also get you into machine learning.
I strongly recommend taking the generalist approach. It makes you a much better engineer in the long run. I had a friend call me up enraged one day. He got woken up at 1am and ordered to drive across the state to get a site back up and running because his hardware was broken and the very expensive software team was sitting around wasting their time. So he got to the site at three or four AM and found that the room full of software engineers, three PhDs among them, didn't understand that BOTH of the battery terminals needed to be connected. Yes, batteries in fact have a positive and a negative. And these people were each getting paid twice as much as him. He left that company and became hardware leadership at a much better company, thank goodness. My point is, super-specializing in the most profitable hyper-niche right off the bat is going to make you a very mediocre robotics engineer who might be successful in the short term, but might find yourself unemployed in a few years when we find a better modality to replace your hyper-niche. Specialists win the sprints, generalists win the marathons.
I've coached a First team that has qualified to go to worlds the last four years and been on the Einstein field once in that time. I think you'd learn a few things - robots are built to solve problems and the most technically complex and advanced robots don't always solve problem better than simple elegant designs. - Robots involve multiple skills and usually get built by teams where members are specialists in specific fields. Some people will understand all of the robot sub systems but
Additionally, find a nearby FRC competition and volunteer for at least one event (do more, if you can). Wander the robot pit and interact with the teams. There will be a lot of good intel for you there just wandering around and asking questions.
Source: Am volunteer judge for FRC.
Just a few thoughts about starting a team and/or volunteering:
Starting a FIRST Robotics Competition (FRC) team is not for the feint-of-heart and requires significant time and money investment. This is the sort of endeavor you try when you're either retired or have a group of other dedicated volunteers willing to help you build the team over multiple seasons. If you spend a year or two mentoring/volunteering for a well-established team, you'll get a good sense of what you're getting into. FRC is a bit hardcore.
If no FRC teams exist nearby, FIRST Lego League (FLL) is a good entrypoint, but may not be technical enough for an adult interested in using FIRST to gain exposure to robotics. (It is plenty satisfying to mentor these teams, but Mindstorms can be somewhat limiting.) These are geared toward using Lego Mindstorms for learning robot concepts.
A good middle ground (for exposure to more practical robotics) would be volunteering with a FIRST Tech Challenge (FTC) team. It's the middle ground between FLL and FRC and is _doable_ for a single, dedicated, passionate adult to start (though I'd strongly recommend finding another volunteer to pursue this with).
If getting involved with a team is not an option, seek out nearby FIRST events that you can volunteer at. Many competition events need volunteers and you'll get a chance to interact with local teams and find plenty of opportunities to play with robots.
For starter, I always recommend a RC car with an Arduino (or cheaper equivalents).
Following can be had for less than <$100
- SG90 servo motor - 28BYJ-48 stepper motor - ULN2003 driver board - UNO R3 Controller Board (Elegoo) - RC car chasis/3D printed chasis
Hack around, add sensors, cameras etc. Next would be to build a robotic arm.
Crunch labs: https://www.crunchlabs.com/
As someone in the industry (I literally can call myself a Robot Master having a Masters Degree in Robotics), these simple “kid kits” are phenomenal. When you first start out, the hardest part is knowing the hardware is actually going to do what you tell it to do. Simplifying the mechanical engineering/electrical engineering so you can immediately create real world actions makes a huge difference in giving you motivation to keep going.
Once you have the basics in place and you understand mechanisms, sensors, and some algorithms, you can move on to other parts of robotics.
However, keep in mind robotics is really about system thinking… you have to understand trades between sensing, thinking, and acting. But most importantly you have to remember you are dealing with the real world… you can’t as easily write a unit test and know that a particular actuator will behave exactly the same way every time.
Just a few weeks ago, along with my partner i started this Sunday experiment of building and learning robotics as a software engineer.
We ordered a one hand robotic arm with orin nvidia gpu to start running experiments.
to commit and make it a habit, we also started documenting our learnings and experience.. if this motivates you you can check it out here - https://youtu.be/pAwkl6zd0ws