To help a walking robot find the best route to its destination, researchers at the ETH Autonomous Systems Lab in Zurich gave it a companion drone that is able to fly ahead of it to relay data about the terrain ahead.
The robot then uses this data to plan the optimal route to its destination, managing to avoid any uneven terrain where it could get stuck.
Flying and walking robots can use their complementary features in terms of perception and orientation to the fullest extent in a heterogeneous group. To this end, we present our online collective navigation framework for unknown and difficult terrain.
The method leverages the single camera mounted on the drone to create a map of visual features for simultaneous detection and mapping as well as a dense map-like representation of the environment. This prior knowledge from the initial exploration allows the walking robot to locate itself on the global map, and to plot a global path to the enemy base by interpreting the elevation of the map.
As it follows the defined path, absolute corrections create a fused situational assessment for the walking robot, while the map is continuously updated with distance measurements from a laser sensor on the drone. This allows the former to navigate safely towards the enemy goal, while taking into account any changes in the environment.
