Democratizing Robotics

In this talk

1. 0:54

   Why robots struggle outside factories

   Industrial robots can move precisely and quickly, but safety fences keep people away. Home robots such as vacuum cleaners still struggle with everyday obstacles, while Predator drones follow human commands rather than operating autonomously. To serve people in places like homes and hospitals, robots must handle changing environments safely and make decisions on their own.

2. 4:22

   Cheaper components make agile robots possible

   Kazan points to falling costs and improvements in motors, sensors, and onboard computing. He compares modern drone motors, which cost about $100 each, with a steam engine, citing a power-to-volume improvement of more than a thousand times. A sensor that once measured missile acceleration has a modern counterpart costing under 500 yen, and a $10 device can perform 200 million instructions per second.

3. 8:27

   Cloud computing and algorithms extend robot capabilities

   Kazan describes cloud computing as rented computing power that robots can use without carrying powerful computers onboard. In a Zurich office demonstration, two low-cost robots sent sensor data to Amazon data centers for 3D mapping, then combined their maps. He also points to ImageNet, with about 10 million images, as data that helped improve object recognition until an algorithm surpassed human performance in 2015.

4. 10:47

   A platform could simplify robotics development

   A European Union hospital-room project required 15 PhDs and sleepless weeks to assemble a one-hour demonstration, despite having a capable team. Kazan argues that businesses need robots to run around the clock with little support, and proposes a platform that abstracts hardware and provides tools to maintain it. This could let hardware makers focus on devices while software developers from fields such as computer vision, machine learning, logistics, and healthcare build applications.

5. 13:17

   Connect developers and robots through a shared platform

   Kazan says robots could use data centers as an extended brain for demanding algorithms and a connected brain for sharing what they learn. System integrators could assemble software building blocks and concentrate on application problems and security details. Re:Peater Robotics is building the platform and seeks software and hardware developers, application builders, and partners.