Improved navigation for androids

Autonomous navigation is one of the most challenging parts of building a robot and especially, if it is a humanoid robot you are dealing with, things easily get much more difficult. Of course, real world conditions involve changing environments as well.

Researchers at Carnegie Mellon University, were able to come up with an algorithm that enables Honda’s famous humanoid robot Asimo to walk through a continuously changing set of obstacles. There is no doubt that doing this task is much more challenging than just navigating through a steady set of obstacles.

The algorithm that the team came up with, called “the footstep planner”, calculates the “cost” of making each possible step among obstacles to the final destination. The cost of a possible step is determined from the estimated travel time and a potential collision with an obstacle. After calculating all possible costs at a given point in time, the robot weighs the risks, and determines the least costing move.

Asimo uses an overhead camera installed at the ceiling of the room, which constantly sees and communicates the changing environment, the position of the robot and the final destination point. With no doubt, the real world applications will most often require the robots to see all these with it own eyes and not with an external camera at a fixed location. Also the obstacles are just papers instead of real objects.

The team says that the research will go further into navigation in more complicated environments such as on a set of steps, or coming up with algorithms for different tasks which involve the use of arms as well. Humanoid robots or androids offer great potential in the future as helper robots as they are the most suited to imitate real human behavior in performing tasks.

See video from YouTube below.