It is getting easier to create intelligent robots, as systems get cheaper and there are new software platforms emerging. A big part of this has been due to the Robot Operating System (ROS), which is becoming a standard in the industry. The origins go back to 2007 when the platform began as an open source project at the Stanford Artificial Intelligence Laboratory.
Despite its name, ROS is really not a true operating system. Instead, it is middleware that helps to manage many of the critical parts of a robot: planning, simulations, mapping, localization, perception, and prototypes. ROS is also modular, as you can easily pick and choose the functions you need. The result is that the system can easily cut down on development time.
Another advantage: ROS has a global community of users. Consider that there are over 3,000 packages for the platform.30
As a testament to the prowess of ROS, Microsoft announced in late 2018 that it would release a version for the Windows operating system. According to the blog post from Lou Amadio, the principal software engineer of Windows IoT, “As robots have advanced, so have the development tools. We see robotics with artificial intelligence as universally accessible technology to augment human abilities.”31
The upshot is that ROS can be used with Visual Studio and there will be connections to the Azure cloud, which includes AI Tools.
OK then, when it comes to developing intelligent robots, there is often a different process than with the typical approach with software-based AI. That is, there not only needs to be a physical device but also a way to test it. Often this is done by using a simulation. Some developers will even start with creating cardboard models, which can be a great way to get a sense of the physical requirements.
But of course, there are also useful virtual simulators, such as MuJoCo, Gazebo, MORSE, and V-REP. These systems use sophisticated 3D graphics to deal with movements and the physics of the real world.
Then how do you create the AI models for robots? Actually, it is little different from the approach with software-based algorithms (as we covered in Chapter 2). But with a robot, there is the advantage that it will continue to collect data from its sensors, which can help evolve the AI.
The cloud is also becoming a critical factor in the development of intelligent robots, as seen with Amazon.com. The company has leveraged its hugely popular AWS platform with a new offering, called AWS RoboMaker. By using this, you can build, test, and deploy robots without much configuration. AWS RoboMaker operates on ROS and also allows the use of services for machine learning, analytics, and monitoring. There are even prebuilt virtual 3D worlds for retail stores, indoor rooms, and race tracks! Then once you are finished with the robot, you can use AWS to develop an over-the-air (OTA) system for secure deployment and periodic updates.
And as should be no surprise, Google is planning on releasing its own robot cloud platform (it’s expected to launch in 2019).32
Leave a Reply