Cloud-based Robot Navigation
Automated guided vehicles (AGVs) are still rigid set-ups in most operating environments where single vehicles follow fixed routes. Their on-board sensors must constantly record the information about their operational environment that they need for path planning. If an obstacle gets in their way, they will stop until the path is clear. Once configured, great effort is required to adapt the AGV to any new layout. To avoid costly and time-consuming conversions for restructuring measures, a four-member research team, led by Group Manager Dr Kai Pfeiffer at Fraunhofer IPA, has linked all the automated guided vehicles and external laser scanners of a workshop together via the cloud.
Local map always up-to-date
One component of cloud navigation is its especially developed cooperative longterm SLAM (Simultaneous Localisation and Mapping) software module: All the permanently installed laser scanners, and the sensors of all the automated guided vehicles, collect information about their surroundings to create a map that is updated continuously. "The cloud-based navigation server calculates the route maps for each individual vehicle from this data," explains Pfeiffer. The 'Predictive Driver' software module, the latest development after 'Elastic-Band', is responsible for reactive path planning. The software responds to spontaneously occurring obstacles and calculates an alternative route. Should the routes of two automated guided vehicles cross, their movement planners will then talk to each other via the cloud, preventing bottlenecks or collisions.
An AGV can be retrofitted with cloud navigation at any time. The cloud solution can be implemented locally, i.e. on-site, as a type of master computer. Initial applications have shown that localisation accuracy increases by as much as 75 percent. Moreover, cooperative path planning reduces travelled distances by up to 20 percent, while smooth traffic at intersections yields time savings of 25 percent. Cloud navigation also produces hardware savings: automated guided vehicles manage with fewer sensors and less computing power because they do not have to use computationally intensive navigation algorithms. "This makes AGVs both more efficient and more economical," summarises Pfeiffer. "Energy consumption per processing unit is reduced by 70 percent, and the cost for sensors by up to 80 percent in certain cases."