4 ms·
Ask HN: What would you like this robot to do?
I have a hunch that the time is right for a robotics revolution (only six years after Bill Gates thought so), and rather than watching it pass me by, I intend to start it. I want to build the company that puts a robot in every home, and I'd like Hacker News to help me find the path to that goal. By way of introduction, I'm an Australian robotics post-doc currently designing and implementing systems to control fleets of mining robots for Rio Tinto.
Phase one of my plan is to build a robot with the following combination of technologies: a Kinect, a laser-projector, an omni-directional mobile platform, and a permanent internet connection. It explicitly does not have any manipulators because I don't believe the tech is suitable or cheap enough for consumers yet. Likewise legs, anthropomorphism, and the idea of robot-as-household-slave. Targeting a price point and market similar to a high-end laptop, I want to ensure the tech is rock solid, sufficiently cheap, and - in combination - becomes a platform technology for subsequent ideas.
The Kinect lets us build maps and localise using SLAM, perform face-tracking and recognition using the camera, and enables voice control using the microphone. The projector lets us draw on any surface (laser projectors are focus-less), and the omni-directional platform lets us keep an image in a stable position while moving the robot elsewhere. The permanent internet connection let us farm out computationally intensive tasks (like learning, inference, and map correction) to the cloud, interact with other internet connected devices in the home, and to source exteroceptive sensory data like news, weather, remote commands, etc. There's a few other features that are less interesting but necessary, like a ring of IR sensors on the base for short-range obstacle avoidance, a pan-tilt head for the projector that lets us point it somewhere different to the Kinect, and batteries sufficient for at least two hours of full activity. To build a business starting from this product, it will eventually require an app store that offers new applications and tasks, as well as new drivers to control or receive information from additional devices.
Ultimately, I envisage this and subsequent robots as the heart of an interconnected home. Like a butler with no arms, but who uses the Force to control all your other electronic devices. (OK, that analogy could use some work...)
My question to HN is - what would you like this robot to do?
Some examples to get the ball rolling:
* Project a ticker-tape of HN headlines on your wall.
* Show previews of the shows on other channels onto the space around your TV.
* Respond to queries such as "Is Sophie in the Lounge Room?" - both in person, and remotely via the web.
* Connect to wifi-enabled lighting and turn it off when people leave the room.
* Monitor and log resident's locations and activities over time, ultimately to enable applications such as optimising heating/cooling in each room.
* Verbally describe your agenda for the day.
* Send you an SMS/email if someone enters your home during the day, with video.
* Animate a school of fish swimming across your walls, avoiding obstacles such as pictures and doorways.
* Using the SLAM map, calculate the dimensions of the house, carpeted area, etc.
As for the next steps, I'm posting this question because the biggest risk factor right now is "does anyone want this?" I work with a lot of talented people, many of them wanting to work on cool consumer technologies such as this, but most too risk adverse to quit their salaries and dive in to a pre-revenue startup. I need to find some partners - preferably two; one hardware, one a better programmer than me - and a mentor who's built a product company before. I've got leads in both directions, but nothing solid at this point. Assuming there's some interest in the concept (and prior to this question, my chats at networking events have all been enthusiastic, but there's a huge selection bias there, and here for that matter) I'm going to run a hackathon with a select few workmates and students on my poaching list. We'll start by building the sensor/projector part, because our laboratory already has the perfect omni-directional platform, and we can probably license that tech. Then, with something to show off, I'll start pitching for grants and investment.
- tgflynn 15y agoI think manipulators are key to the vast majority of truly useful robotic applications. I'd be very interested to know why you think they are so difficult. Why would controlling arms be so march harder than controlling a self-driving car in San Francisco.
- Schwolop 15y agoIts not so much that I think they're difficult (although, they really are!), rather that they dramatically increase the computational power required for planning, the sensing required for pose estimation and object tracking, etc. Take Willow Garage's PR2 for example, it's a great robot that can do mobile manipulation, but there's no way a consumer is going to pay $285k (or $400k for the two-arm version).
- tgflynn 15y agoBut why does it need to cost $400K ? Processor cores aren't that expensive.
- Schwolop 15y agoNo, but they increase heat, energy use, weight, etc. Then you need more batteries, better cooling, bigger motors to carry the weight, etc. To do robust pose estimation of arbitrary objects (cups, books, shirts, crumpled up pieces of really important paper, etc) is an unsolved problem, and current solutions require very expensive sensors. In particular, to accurately grab something requires extremely good localisation, which means you need a truly robust SLAM implementation, or more likely, a better IMU. Getting 2-5X better in IMU terms means roughly two orders of magnitude in price. I absolutely want to build a mobile manipulator robot one day, but I'm 99.9% convinced that it shouldn't be the first thing I build.
- tgflynn 15y agoSo it seems it breaks down into 2 distinct issues : - cost of the arm mechanics - expensive sensors because the algorithms aren't yet good enough to do what's needed with cheap sensors I don't know much about the sensor issues but I'm unaware of any fundamental reason why an arm which basically consists of some metal supports, joints and a handful of smallish motors should cost more than a new car. As for the algorithm issues - humans seem to be able to make do with 2 eyes which seem about equivalent to a pair of cameras which aren't expensive at all. I take that as a proof of principle that the problem is solvable. Looking at how much progress has been made on robotic cars since the first DARPA challenge my bet is that a well focused effort could bring us close to a practical CV based solution within a couple of years.