It is interesting that this robot uses the Kinect camera system rather than the complex sensors used in the past. It seems that as we continuously develop vision processing that it becomes more useful. It is also often less expensive. Sometimes it is very inexpensive. I have an older BlackBerry Curve. It uses a trackball. I have replaced the trackball. It cost about $2.50. Newer models use a low resolution camera in place of the trackball. It only has to sense the direction of movement, not any other details. So, it works fine and is longer lived than the trackball. It is also simpler to build and probably cheaper to install. Any software cost is amortized over all the devices sold, so that is near zero. This is the same with the robot.
I have seen the robots with multiple laser sensors and sonar or radar. These were fantastically expensive and still not as good as a human operator. Humans use vision. Perhaps the MIT researchers are on to something here.
naperlou, I was also interested to see the Kinect motion sensing camera/system used in aiding with 3D mapmaking. To me, when I read this it was one of those "of course" moments. The team also used laser scanners in a previous rev of this project.
I've been hearing a lot about the Kinect motion system in simulation and other types of 3D apps as well. It seems like another one of those instances where consumer technology is influencing the development of commercial/business applications, which is interesting.
What's been happening for several years is all kinds of apps and industries leveraging the high volumes--and therefore relatively low prices--of off-the-shelf consumer and commercial hardware and software.This has certainly operated in the military for awhile now, and has begun influencing machine vision, and to some extent now automation & control, more recently. The pace has really picked up recently with the convergence of multiple technologies at the chip and board levels--witness Kinect--and with multi-core processors.
TJ, I agree with you about the unexpected uses of technology innovations. In the case of the Kinect camera/sensor system for this robot app, note that the team also implemented their approach with the Kinect system in a robotic wheelchair and a portable sensor suit, in addition to the PR2 platform.
What I'm getting from your recent spate of apps stories, Ann, is that robotics apps are extending their arms, so to speak, well beyond the straight industrial arena in which I already assumed they were in heavy use. But there seem to be numerous medical, mil, and other apps of which I'd been unaware. Very interesting.
Thanks, Alex. Yes, I was intrigued to discover that robots are all over the place: they're not just for industry anymore, although that's their largest area of concentration. They are in medicine and healthcare, outer space, used by the military in the air and on the ground, and are learning to do all kinds of new things like navigate autonomously in a novel environment, fly in formation in swarms, build structures and even play in concert:
Chuck, that video went viral in about a week after it was posted. Although I think much of that is due to the cute and/or novelty factor, I also think much of the engineering appeal will be wondering how they are synchronized in yet another form of swarming behavior.
This is interesting TJ, the ethics question came to me also, although I was thinking more during war time. I think we have all seen videos of enemy soldiers trying to surrender to drones - before they get blown up.
The influx of robots in war raises a unique moral question about surrendering to a mechanical entity that a human is monitoring.
ChasChas, that's an interesting question you pose. But some of these newer robots will be functioning autonomously, like this one, i.e., not under direct human control. So if these are designed as soldiers, not as merely explorers, the ethical situation changes somewhat.
Ann, A similar industrial application is mobile robots designed to integrate into warehouse management , scheduling and inventory control systems. The robots are designed to do sophisticated, autonomous behaviors, navigation and localization without having a robot assist or programmer involved by using natural, features-based autonomous navigation and localization to eliminate the need to install lines or beacons used with other AGV solutions. The robots learn their location in a facility based on natural features, learning the facility layout by looking at walls, parts of the factory or the ceiling in a dynamic space like a warehouse. Because they don't require beacons or lines in the floor, it reduces the upfront cost of the system. I believe the systems are used to transport finished auto tires in large warehouse facilities. Definitely lots of innovations in robotics that go beyond traditional applications.
was my first encounter with what are called autonomous robots, and the one in this story is my second. Both made me wonder where else that idea is being used, and what different technologies make them possible, in particular, the navigation and map-making abilities. The industrial environment is certainly an obvious choice.
Ann, The key technology with the mobile robots I've seen is software enhancements and intelligent algorithms. Enhancements in vision systems, for example, provides the mechanism to visualize and ultimately "map" the factory environment but in the end the most difficult task is the mass of intelligent software required. It ranges from becoming an expert system (gathering information to make more informed decisions) to advanced databases for storing information. Lots of software
One additional area of software innovation for mobile robots is algorithms for obstacle avoidance. Especially in systems where the mobile robot will encounter humans, such as the tire warehousing application, where the robot is "delivering" a completed tire to a storage/retrieval system, the mobile robot can encounter workers during that delivery process. The software to control those interactions are interesting and also critical to the success of the application.
Al, I agree that obstacle-avoidance and mapmaking software is a big deal. Specifically, the map-making/obstacle avoidance algorithms based on Simultaneous Localization and Mapping (SLAM) techniques mentioned here, which may also be what's behind the tiny swarming robots' mapmaking ability:
Ann, Used as a tool to aid in developing mobile robots, the Kinect sensor provides a unique type of feedback which can be used in conjunction with flexible I/O, software algorithms and real-time controllers to quickly and easily prototype, test and deploy robotic applications. The development tools already available make it great for prototyping. But while the Kinect is useful for common robot tasks such as obstacle avoidance, like most sensors it also has limitations. For example, the Kinect cannot detect obstacles that are closer than two feet and does not work well in the sunlight. Still great technology at a mind boggling cost.
Al, thanks for those additional details on the Kinect sensor's limitations. The fact that it doesn't detect objects less than two feet away should not be a deterrence to its use in checking out a new environment for military tasks, such as in advance of first responders. But I'm surprised that it doesn't work well in sunlight--that seems like a major limitation for these applications, and for helping the elderly or disabled, both of which were two applications the MIT team mentions and which occur at least partly in sunlight. I would not be surprised if this research team is working on methods for overcoming that problem, also.
For industrial control applications, or even a simple assembly line, that machine can go almost 24/7 without a break. But what happens when the task is a little more complex? That’s where the “smart” machine would come in. The smart machine is one that has some simple (or complex in some cases) processing capability to be able to adapt to changing conditions. Such machines are suited for a host of applications, including automotive, aerospace, defense, medical, computers and electronics, telecommunications, consumer goods, and so on. This discussion will examine what’s possible with smart machines, and what tradeoffs need to be made to implement such a solution.