Although robots are designed by people, getting robots to perform useful tasks in a safe manner is an ongoing challenge. In order to teach robots how to perform their desired tasks, researchers like SFU computing science professor Mo Chen design artificial intelligence (AI) algorithms to satisfy performance and safety requirements.
“For deep learning-based AI, we try to incorporate prior knowledge about the robots and their environments into learning algorithms so that robots can learn to perform their tasks more quickly,” says Chen.
Finding ways to integrate useful and safe robots into society in a way that assists people is a process that involves working on theory and simulations before experimenting with actual robots. Chen directs the Multi-Agent Robotic Systems Lab at SFU, where he and his team place an emphasis on fundamental, theoretical foundations, in addition to real world implementation.
“Theory is very important for designing algorithms that are guaranteed to be safe and for making learning algorithms more efficient,” says Chen. “To test our theory, we need to do a lot of simulations to make sure that the algorithm is doing what it should be doing. The next challenge is taking what the robots have done in simulations and transferring it to the real world. A lot of times we find that there are differences in the real world from simulation, so we often have to go back and refine what we developed.”
While this process is long and challenging, Chen focuses on the good that this research can do in the future. He sees many ways that robots can have a positive impact on society, whether it is having a robot greet you and help you find items in a store, using drones to help film movies or using robots to gather agricultural data to help people understand the well-being of crops.
For these goals to become a reality, however, there is still a lot of research and testing that needs to be done. Beyond the challenge of designing robots to complete desired tasks, Chen focuses on ensuring that robots are safe while maintaining their usefulness so that people can trust robots. As time goes on, he believes that we will see robots gradually integrate into society as people become more comfortable with them.
For the future, Chen is excited to continue collaborating with other researchers. In addition to his current research, he hopes to train robots to understand their effects on the environment and other agents, and make decisions based on these effects. This will help robots interact more naturally and safely with people.