At this point, whenever I see a Boston Dynamics youtube video, I mentally prepare myself for some uncanny-valley robot action. Ever since I watched that dog robot get kicked on the ice in 2010(!),...
At this point, whenever I see a Boston Dynamics youtube video, I mentally prepare myself for some uncanny-valley robot action. Ever since I watched that dog robot get kicked on the ice in 2010(!), I have been patiently watching the developments of the robot world.
I wonder how long it will be until the use of personal robot helpers comes to fruition. Imagine sending your robot to drop off your packages or prepare your lunch. 2030? 2040?
I can see the applicability of robots in distribution and retail environments, but I'm most curious about what it will be like when the Roombas become more than just a vacuum cleaner.
Trust me, it'll be sooner than that. Robotics are finally reaching their exponential possibilities due to knowledge, pricing, manufacturing, and advances in AI and general processing power....
I wonder how long it will be until the use of personal robot helpers comes to fruition. Imagine sending your robot to drop off your packages or prepare your lunch. 2030? 2040?
Trust me, it'll be sooner than that. Robotics are finally reaching their exponential possibilities due to knowledge, pricing, manufacturing, and advances in AI and general processing power.
(Source: I am a roboticist and founder of a robotics company that works on very similar developments).
The biggest hurdle right now is battery life though, isn't it? E.g. IIRC, Spot (the yellow dog-like robot) only has 90min of battery life... if operating on totally flat terrain without carrying...
The biggest hurdle right now is battery life though, isn't it? E.g. IIRC, Spot (the yellow dog-like robot) only has 90min of battery life... if operating on totally flat terrain without carrying any significant load. For commercial/industrial use, battery life probably isn't really much of an issue since those robots can just stay tethered to power lines, but for personal/household use that short of a battery life would significantly limit their usefulness, IMO.
Most think this! With lithium ion and the motors (most are DC brushless) then the weight and life of the battery is fairly minimal. A single 24v 20aH battery (a car battery) will be able to move...
Most think this! With lithium ion and the motors (most are DC brushless) then the weight and life of the battery is fairly minimal. A single 24v 20aH battery (a car battery) will be able to move 3-600lbs at 3mph for about 8-9 hours without a charge if you have a 250W motor (lowest standard in these guys). Those are heavy (Sealed Lead Acid - 24lbs) but there are now Lithium Ion 24v 20aH batteries coming in at around 3lbs. That is the same capacity, with the same amount of power at an eighth of the weight and a smaller form factor. Side note: Recharging a full battery of that caliber off a standard power outlet is only about 8 hours or so.
Battery life is a major obstacle in cars that have to go hundreds of miles at sustained 40+ mph without a charge, but for delivery bots, construction, warehouses, around the house, you only need a few miles of range and a far lower speed. As for price, the Lithium-Ion batteries are 4x the cost, coming in at around $140 at a general market ($70-80 in wholesale)
edits: cleared up some generalization, added details, and added a few more generalizations.
Perhaps we’re getting a misleading impression from drones, which spend lot of power staying in the air? Meanwhile, Boston Dynamics has put a lot of effort into legged robots. This seems like a...
Perhaps we’re getting a misleading impression from drones, which spend lot of power staying in the air?
Meanwhile, Boston Dynamics has put a lot of effort into legged robots. This seems like a pretty big power drain? Is this actually a promising direction compared to wheels?
Most of the BD robots use hydraulics instead of standard motor drivers, this is great for strength but has a lot of issues as it requires major maintenance and the power drain is indeed bonkers,...
Most of the BD robots use hydraulics instead of standard motor drivers, this is great for strength but has a lot of issues as it requires major maintenance and the power drain is indeed bonkers, along with the weight.
They aren't the only ones working on legged robots, but when a serious industrial wheeled robot that can lift loads of multi-thousands and run for several hours is $80k and then you have the legged Spot that just can't do either of those and runs for only 90m is also $80k you can see where the issue of power and price comes into play.
The prospect is exciting, but I am bullish on their commercial endeavors. Now, the individual parts (not the whole robot) are going to be groundbreaking once they start selling them to be used in other robots. They have a lot of really cool innovative individual parts.
Ahhhh, okay. Suddenly it all clicked into place. I was wondering how you were talking about great battery life when all the BD bots have such short ones. I didn't realize they were using a much...
Ahhhh, okay. Suddenly it all clicked into place. I was wondering how you were talking about great battery life when all the BD bots have such short ones. I didn't realize they were using a much more power hungry hydraulic system for theirs. Thanks for the insight!
If you take a look at the mat in the video, it is covered in hydraulic fluid. Close to the end you can see puddles of it all over, looks like Atlas sprung a leak! Hydraulics are superior for both...
If you take a look at the mat in the video, it is covered in hydraulic fluid. Close to the end you can see puddles of it all over, looks like Atlas sprung a leak!
Hydraulics are superior for both speed and strength, but costly in terms of raw power and energy consumption.
I'd really like to read some papers that make you so optimistic (yours or other peoples'). I'm not in robotics directly, but I work adjacently and have a few friends who are PhDs working actively...
I'd really like to read some papers that make you so optimistic (yours or other peoples'). I'm not in robotics directly, but I work adjacently and have a few friends who are PhDs working actively in the field. So far I see a ton of money being poured into much simpler robotic platforms (specifically self-driving cars and "Urban Air Mobility" self-flying drones) with virtually no progress over the last 5 or so years.
The autonomy problem space for autonomous vehicles seems extremely constrained compared to the problem space that a android-like robot has to deal with.
It may seem like there's not been any major developments in the last five years, but believe me when I say we have more innovation and progress in the last five years than we have in the previous...
It may seem like there's not been any major developments in the last five years, but believe me when I say we have more innovation and progress in the last five years than we have in the previous forty years of robotics or automation.
I don't have access to any major papers at the moment (not at work, nor home) but Rodney Brooks has written about the Cambrian explosion of robotics for about twenty years and has been amazingly accurate in his time lines. Here is a paper from 2015 that is more true now than it was then.
That's exciting! Do you have any recommendations on how to get started as a hobbyist in this field? I have looked into kits and information in the past, but it has often been a bit overwhelming.
That's exciting!
(Source: I am a roboticist and founder of a robotics company that works on very similar developments).
Do you have any recommendations on how to get started as a hobbyist in this field? I have looked into kits and information in the past, but it has often been a bit overwhelming.
This is totally common place and stereotypical, but Arduino! You can get a lot of the basics down really quickly, low cost, and all you need is a PC, an Arduino, a breadboard, and a few servos!...
Exemplary
This is totally common place and stereotypical, but Arduino!
You can get a lot of the basics down really quickly, low cost, and all you need is a PC, an Arduino, a breadboard, and a few servos! From there you can build a lot of very cool and interesting (and sometimes useful!) stuff!
I also like messing with vibration motors with mine, in fact I hate using LED's on them so I use that for Hello Worlds :)
There has never been a better and more beautiful time to get into electronics or robotics! Keep in mind if an idea requires a 3D printer (and you don't have access to one), you can usually get by with some hot glue and craft sticks (tongue depressors), also have fun and find things around the house, I use a lot of random plastic containers to build shapes for fun DIY projects, and cardboard and duct tape is always fantastic. Don't let anything stop you from building and having fun with it!
It will surely be after self-driving cars become common. Self driving cars are a much simpler problem to solve. The space is far more constrained and predictable, and yet there's little evidence...
I wonder how long it will be until the use of personal robot helpers comes to fruition. Imagine sending your robot to drop off your packages or prepare your lunch. 2030? 2040?
It will surely be after self-driving cars become common. Self driving cars are a much simpler problem to solve. The space is far more constrained and predictable, and yet there's little evidence that we'll see self-driving cars that can handle most weather conditions within the next decade or two.
Snow is far from the only unsolved problem for self-driving cars. Computer "vision" (in visible and other wavelengths) just isn't even remotely close to the level needed for mass adoption of...
Snow is far from the only unsolved problem for self-driving cars. Computer "vision" (in visible and other wavelengths) just isn't even remotely close to the level needed for mass adoption of unattended self-driving vehicles in a wide range of applications. A useful personal robot has far greater needs in the vision department. It needs to be able to interact with things on a much finer scale than a car.
I shouldn't have said "weather conditions" original. Just "conditions" would have been more accurate. Weather is part of it, but in general computer visions systems are good at recognizing a car that looks like a car and road markings that looks like well-maintained road markings, but they have some very bad failure cases when presented with conditions outside the norm. If your personal robot walks out into traffic because it misjudges conditions that's more than an inconvenience.
Not exactly. A personal robot will have fine tuned motor skills and only require paying attention to a much smaller area and only doing one or two tasks at a time (scanning for an item [1],...
personal robot has far greater needs in the vision department. It needs to be able to interact with things on a much finer scale than a car.
Not exactly. A personal robot will have fine tuned motor skills and only require paying attention to a much smaller area and only doing one or two tasks at a time (scanning for an item [1], grabbing an item[1], transporting an item [2, holding and moving] ), where a car requires tracking everything within a 30m area around, at high speed, with a thousand different unknown variables all rapidly changing. A personal robot is more stationary, requires less analysis of an area, and will be more focused on a single task at hand.
Now for the purposes of pedestrian traffic this is a problem due to the different types of areas. If you should take a road or side walk, navigating around pot holes, going up and down curbs. But there are some great new tech that's starting to solve that. Delivery bots are a very hot thing right now and having quite great success in small markets.
A personal robot has far less vision and safety requirements than a high speed, high load, variable changing, steel behemoth.
The visual requirements for finite small details for interacting with things is a lot larger, but these are far easier to manipulate than they were even five years ago.
These are human-scale issues, and only marginally dangerous.. People do both those dumb things all the time, and we have ways of handling that. A human sized thing falling over or stepping into...
These are human-scale issues, and only marginally dangerous.. People do both those dumb things all the time, and we have ways of handling that. A human sized thing falling over or stepping into traffic is much less dangerous than a car driving in ways that have no respect for human life.
Human scale issues aren't a problem (a big problem, that is) precisely because its human scale. It ends at any human and next human likely doesn't have that issue. The semi-random nature of human...
Human scale issues aren't a problem (a big problem, that is) precisely because its human scale. It ends at any human and next human likely doesn't have that issue. The semi-random nature of human perception and action ensures that.
A machine driving ALL the cars scales an issue, any issue everywhere, with near perfect replication. That, IMO is the concern.
I thought the same, not because of what the robots are doing but because of the lighting. I think the lighting is just exceptionally even, which lends it to looking a bit like a render.
I thought the same, not because of what the robots are doing but because of the lighting. I think the lighting is just exceptionally even, which lends it to looking a bit like a render.
I don’t think it’s the lighting, but what they’re doing is something my brain already classify as “not possible in real life” and immediately associate with fiction. The video also have a...
I don’t think it’s the lighting, but what they’re doing is something my brain already classify as “not possible in real life” and immediately associate with fiction. The video also have a proto-narrative and convey emotion, something that I associate with movies like Chappie and Real Steel, for example.
At this point, whenever I see a Boston Dynamics youtube video, I mentally prepare myself for some uncanny-valley robot action. Ever since I watched that dog robot get kicked on the ice in 2010(!), I have been patiently watching the developments of the robot world.
I wonder how long it will be until the use of personal robot helpers comes to fruition. Imagine sending your robot to drop off your packages or prepare your lunch. 2030? 2040?
I can see the applicability of robots in distribution and retail environments, but I'm most curious about what it will be like when the Roombas become more than just a vacuum cleaner.
Trust me, it'll be sooner than that. Robotics are finally reaching their exponential possibilities due to knowledge, pricing, manufacturing, and advances in AI and general processing power.
(Source: I am a roboticist and founder of a robotics company that works on very similar developments).
The biggest hurdle right now is battery life though, isn't it? E.g. IIRC, Spot (the yellow dog-like robot) only has 90min of battery life... if operating on totally flat terrain without carrying any significant load. For commercial/industrial use, battery life probably isn't really much of an issue since those robots can just stay tethered to power lines, but for personal/household use that short of a battery life would significantly limit their usefulness, IMO.
Most think this! With lithium ion and the motors (most are DC brushless) then the weight and life of the battery is fairly minimal. A single 24v 20aH battery (a car battery) will be able to move 3-600lbs at 3mph for about 8-9 hours without a charge if you have a 250W motor (lowest standard in these guys). Those are heavy (Sealed Lead Acid - 24lbs) but there are now Lithium Ion 24v 20aH batteries coming in at around 3lbs. That is the same capacity, with the same amount of power at an eighth of the weight and a smaller form factor. Side note: Recharging a full battery of that caliber off a standard power outlet is only about 8 hours or so.
Battery life is a major obstacle in cars that have to go hundreds of miles at sustained 40+ mph without a charge, but for delivery bots, construction, warehouses, around the house, you only need a few miles of range and a far lower speed. As for price, the Lithium-Ion batteries are 4x the cost, coming in at around $140 at a general market ($70-80 in wholesale)
edits: cleared up some generalization, added details, and added a few more generalizations.
Perhaps we’re getting a misleading impression from drones, which spend lot of power staying in the air?
Meanwhile, Boston Dynamics has put a lot of effort into legged robots. This seems like a pretty big power drain? Is this actually a promising direction compared to wheels?
Most of the BD robots use hydraulics instead of standard motor drivers, this is great for strength but has a lot of issues as it requires major maintenance and the power drain is indeed bonkers, along with the weight.
They aren't the only ones working on legged robots, but when a serious industrial wheeled robot that can lift loads of multi-thousands and run for several hours is $80k and then you have the legged Spot that just can't do either of those and runs for only 90m is also $80k you can see where the issue of power and price comes into play.
The prospect is exciting, but I am bullish on their commercial endeavors. Now, the individual parts (not the whole robot) are going to be groundbreaking once they start selling them to be used in other robots. They have a lot of really cool innovative individual parts.
Ahhhh, okay. Suddenly it all clicked into place. I was wondering how you were talking about great battery life when all the BD bots have such short ones. I didn't realize they were using a much more power hungry hydraulic system for theirs. Thanks for the insight!
If you take a look at the mat in the video, it is covered in hydraulic fluid. Close to the end you can see puddles of it all over, looks like Atlas sprung a leak!
Hydraulics are superior for both speed and strength, but costly in terms of raw power and energy consumption.
I'd really like to read some papers that make you so optimistic (yours or other peoples'). I'm not in robotics directly, but I work adjacently and have a few friends who are PhDs working actively in the field. So far I see a ton of money being poured into much simpler robotic platforms (specifically self-driving cars and "Urban Air Mobility" self-flying drones) with virtually no progress over the last 5 or so years.
The autonomy problem space for autonomous vehicles seems extremely constrained compared to the problem space that a android-like robot has to deal with.
It may seem like there's not been any major developments in the last five years, but believe me when I say we have more innovation and progress in the last five years than we have in the previous forty years of robotics or automation.
I don't have access to any major papers at the moment (not at work, nor home) but Rodney Brooks has written about the Cambrian explosion of robotics for about twenty years and has been amazingly accurate in his time lines. Here is a paper from 2015 that is more true now than it was then.
That's exciting!
Do you have any recommendations on how to get started as a hobbyist in this field? I have looked into kits and information in the past, but it has often been a bit overwhelming.
This is totally common place and stereotypical, but Arduino!
You can get a lot of the basics down really quickly, low cost, and all you need is a PC, an Arduino, a breadboard, and a few servos! From there you can build a lot of very cool and interesting (and sometimes useful!) stuff!
I also like messing with vibration motors with mine, in fact I hate using LED's on them so I use that for Hello Worlds :)
I dislike tutorials (I'm a hacker at heart) but there are a lot (589) tutorials/projects on the Arduino Robotics project site, a few thousand on the hackaday.io site, or on Hackster.io with a few thousand more!
There has never been a better and more beautiful time to get into electronics or robotics! Keep in mind if an idea requires a 3D printer (and you don't have access to one), you can usually get by with some hot glue and craft sticks (tongue depressors), also have fun and find things around the house, I use a lot of random plastic containers to build shapes for fun DIY projects, and cardboard and duct tape is always fantastic. Don't let anything stop you from building and having fun with it!
It will surely be after self-driving cars become common. Self driving cars are a much simpler problem to solve. The space is far more constrained and predictable, and yet there's little evidence that we'll see self-driving cars that can handle most weather conditions within the next decade or two.
Snow is far from the only unsolved problem for self-driving cars. Computer "vision" (in visible and other wavelengths) just isn't even remotely close to the level needed for mass adoption of unattended self-driving vehicles in a wide range of applications. A useful personal robot has far greater needs in the vision department. It needs to be able to interact with things on a much finer scale than a car.
I shouldn't have said "weather conditions" original. Just "conditions" would have been more accurate. Weather is part of it, but in general computer visions systems are good at recognizing a car that looks like a car and road markings that looks like well-maintained road markings, but they have some very bad failure cases when presented with conditions outside the norm. If your personal robot walks out into traffic because it misjudges conditions that's more than an inconvenience.
Not exactly. A personal robot will have fine tuned motor skills and only require paying attention to a much smaller area and only doing one or two tasks at a time (scanning for an item [1], grabbing an item[1], transporting an item [2, holding and moving] ), where a car requires tracking everything within a 30m area around, at high speed, with a thousand different unknown variables all rapidly changing. A personal robot is more stationary, requires less analysis of an area, and will be more focused on a single task at hand.
Now for the purposes of pedestrian traffic this is a problem due to the different types of areas. If you should take a road or side walk, navigating around pot holes, going up and down curbs. But there are some great new tech that's starting to solve that. Delivery bots are a very hot thing right now and having quite great success in small markets.
A personal robot has far less vision and safety requirements than a high speed, high load, variable changing, steel behemoth.
The visual requirements for finite small details for interacting with things is a lot larger, but these are far easier to manipulate than they were even five years ago.
Until it stumbles and falls on some old person, or invades the road incorrectly and causes a car crash.
These are human-scale issues, and only marginally dangerous.. People do both those dumb things all the time, and we have ways of handling that. A human sized thing falling over or stepping into traffic is much less dangerous than a car driving in ways that have no respect for human life.
Human scale issues aren't a problem (a big problem, that is) precisely because its human scale. It ends at any human and next human likely doesn't have that issue. The semi-random nature of human perception and action ensures that.
A machine driving ALL the cars scales an issue, any issue everywhere, with near perfect replication. That, IMO is the concern.
Somehow it all looks like CGI to me. I know it's real but it looks so unreal.
I thought the same, not because of what the robots are doing but because of the lighting. I think the lighting is just exceptionally even, which lends it to looking a bit like a render.
I don’t think it’s the lighting, but what they’re doing is something my brain already classify as “not possible in real life” and immediately associate with fiction. The video also have a proto-narrative and convey emotion, something that I associate with movies like Chappie and Real Steel, for example.
"Any sufficiently advanced technology is indistinguishable from magic." - Arthur C. Clarke
Man. Robot dance numbers are gonna be like the tool assisted speedruns of real life.