Robotics and automation are already a cornerstone of commercial manufacturing. Predictable environments and repetitive actions make it relatively simple for programmers to teach robots how to behave and perform tasks safely.
In the real world, however, robots will need to be adaptive. Angela Schoellig, assistant professor at the University of Toronto Institute for Aerospace Studies (UTIAS), builds smart robots to extend human capabilities, much like how computers and the internet have changed our daily lives, bringing robots out of the factory and into the real world alongside people.
Schoellig believes that “the biggest challenge that we are addressing is that those robots have to cope with unknown environments, unpredictable situations. Our research is really concerned with enabling them to learn by themselves based on their experience.”
Take, for instance, a self-driving car. Aside from avoiding collisions with pedestrians or other cars, the road and weather conditions are constantly changing. Sunshine and snowfall change the environment drastically. To operate safely, these cars could be programmed to drive very conservatively all the time, but this would not be ideal. With adaptive programming a robotic car could drive conservatively for a few moments while assessing the road, and then follow the program that best suits its current conditions.
Schoellig’s research takes these principles even further, allowing robots to share information and learn from one another, even if those robots are not identical. Her YouTube channel is filled with videos of robots learning and adapting, and even dancing to music as they fly. Her robots can also offload heavy computing onto the cloud, using the power of the internet to help them make decisions.