When they venture out, the current generation of walking robots are provided with existing map data, which can be months and even years old depending on the source. Any new terrain information is gathered in real-time using onboard cameras, which are capable of detecting items in the robot’s immediate surroundings. As a result, a walking robot must scan its local environment and make changes on the fly. It cannot plan its route of travel too far in advance because it cannot detect potential obstacles.
The system designed by Autonomous Systems Lab takes the burden of terrain detection off the robot and places it onto a drone, which acts as scout. In their experiments, a camera-equipped hexacopter drone would fly ahead of the walking robot, mapping the terrain, recording landmarks and calculating the elevation as it goes along. This up-to-date information then was sent to the robot, which combined this local data with its global map to choose the best path for traveling across its terrain. This information is continuously updated by the drone, allowing the robot to receive a steady stream of navigational data that it can use to make course adjustments as necessary.
In the experimental course, the walking robot was able to successfully navigate previously unknown terrain using the map data provided by the drone. It walked without incident from its starting point to the goal position, traversing narrow pathways, scaffolding and even a ramp along the way. These experiments are part of a broader initiative to develop the next generation robotic quadrupeds that will be energy efficient well as capable of dynamic motions including a fast walk, trot, bound, and gallop.
