Researchers at Carnegie Mellon University have developed novel approaches to humanoid robot navigation and path planning using a Honda Asimo and HRP-2 Promet as their test subjects. The robots are able to perceive obstacles in their environment and walk around them to reach a goal destination. They can even predict the velocity of moving obstacles and time their footsteps in order to get through unharmed using computer vision algorithms.
Turning AIST’s robotics laboratory in Japan into a makeshift motion-capture studio allowed them to augment reality by matching real-world objects, such as tables and chairs, with 3d models. This allowed the HRP-2 Promet humanoid to see obstacles in its environment both in simulation and the real world. Two incredible videos after the break.