Wednesday February 15, 2017

Microsoft Shares Open Source System for Training Drones and Other Gadgets

Microsoft has started an open source project for creating smarter drone A.I.. These open source A.I. routines will make it so that anyone can assist in testing the software within a simulator; and then implement it into their own drones. The goal of the project is to get drones to be able to differentiate between a shadow and a pole in the real world for example. They need to know how to avoid people and stop for objects in their way like trees. The further objective for the project is to create devices that can do the laundry, drive us to the store, deliver packages, and other tasks in the real world. Most A.I. today focuses on doing tasks in controlled environments like playing chess for example. Those machines can't operate in the real world because they lack a set of rules to go by to exist with humans and obstacles such as walls.

Another really interesting aspect of the project is the reliance on big advances in graphics hardware, computing power, and algorithms. Aerial Informatics and Robotics Platform software runs the simulator and allows these technologies to be used to assist in the simulation. This allows researchers to crash a drone hundreds of times in a simulation without spending money on drone parts. "If you really want to do this high-fidelity perception work, you have to render the scene in very realistic detail آ– you have sun shining in your eyes, water on the street," said Shital Shah, a principal research software development engineer who has been a key developer of the simulator. These powerful tools also allow for reinforcement training at a faster rate.

News Image

In addition to the simulator, the Aerial Informatics and Robotics Platform includes a library of software that allows developers to quickly write code to control drones built on two of the most popular platforms: DJI and MavLink. Normally, developers would have to spend time learning these separate APIs and write separate code for each platform.

The researchers expect to add more tools to the platform down the road, and in the meantime they hope that the library and simulator will help push the entire field forward.