This week I attended an exciting seminar by Marc Raibert. He is a former MIT Professor who founded the Boston Dynamics Corporation in 1992 as a spin-off from MIT. The company develops a quadruped robot called BigDog among other types of innovative robots, including PETMAN, an anthropomorphic robot for testing equipment, RISE, a robot that climbs vertical surfaces, SquishBot, a shape-changing chemical robot that moves through tight space, and etc.
The BigDog is different from other robots in that it’s designed to operate in rough terrain like rocky, muddy, sandy and snowy surfaces. It can walk, trot, jog, climb a slope, follow a person, and even dance. Marc showed several cool videos, some of which are actually available on Youtube as well.
Bothered by SLOW Web UI to manage vSphere? Want to manage ALL your VMware vCenters, AWS, Azure, Openstack, container behind a SINGLE pane of glass? Want to search, analyze, report, visualize VMs, hosts, networks, datastores, events as easily as Google the Web? Find out more about vSearch 3.0: the search engine for all your private and public clouds.
As you may have known, robots have many real world use cases. For example, it can help to carry weapons in battlefields, move heavy logistics for exploring wild areas, etc. In daily life, it can be a house maid who can help to handle house chores, take care of your kids; it can be a replacement for the Segway.
Somehow the robot technology hasn’t moved forward as much as other technologies like Internet (we passed Web 2.0 and are moving cloud nowadays. ). When I saw the BigDog videos, I said to myself: “maybe it’s a turning point in the robot technology and its applications.”
The robot technology is very hard because it’s a combination of many different technologies, the far most of which is the mechanics including the structure, actuator, and engine. For the structure, the platform, hips, knees, ankles, feet have to be able to work together seamlessly.
Secondly, you have sensors all over the places so that you can collect enough information for controlling different parts so that it can balance and move.
Thirdly, the control part. It could be the hardest when the former two are there. It of course needs computer programming. It turns our programming is not the crucial part; modeling the movement is.
Basically the model should be able to decide quickly what each actuator should do when the sensors collect status. The complexity grows dramatically when you have many sensors detecting different parameters like location, speed, acceleration, and etc. You may be wonder if neural network helps. Maybe, but Marc said they used zero of neural network.
The control part is similar to that of rocket science although it’s not the rocket science. It will only get more sophisticated in the future.
Looking forward, Marc sees key challenges in these areas:
- Energy efficiency so that it can move longer distance
- Leg/body architecture
- Noise reduction
- Terrain perception
After solving these problems, I think the BigDog will be practical in many cases in real world. As any other emerging technologies, it needs use cases to drive it to wide adoption. That is challenging as well. So the business development for killer applications is also critical for the technology.
Now is there any way for the BigDog to relate to the cloud computing? Not until it’s networked with wireless. Once it’s there, it can, for example, leverages Google Maps to find a shortcut to its destination. Also, the BigDog-in-chief could live in the cloud where it can coordinate its subordinates to do something otherwise impossible. That is probably something we can call BigDog 2.0, based on our IT naming convention.