In the last two months of TRAFCON, experiments for comparing the TRAFCON traffic manager and the traffic manager currently used by E80 have been carried on. Experiments have been done in the arena using three AGVs and two different layouts. The experiments have shown that the TRAFCON strategy works better than the E80 one from two main points of view. First, using TRAFCON more missions per hour can be executed; in the worst case roughly the same number of missions achieved by the E80 strategy are executed. Second, it is completely automatic, while the E80 strategy requires a lot of rules to be introduced manually.

In the last two months the experimental phase of the traffic control strategy developed in Task 2 in the arena built during Task 3 has been started. Given the size of the arena, 3 AGVs will be used to emulate realistic traffic situations. Several layouts and the corresponding roadmaps have been designed and the safety regions for the AGVs have been tuned. For each layout a set of missions to be executed by the AGVs on the roadmap has been defined. Several experiments have bee run using three AGVs on various layouts in order to check the presence of some inconsistencies of the traffic manager. The TRAFCON algorithm behaved as expected and a short movie of these experimental results has been uploaded in the Echord website. In the next months, the experimental campaign will be run. The efficiency of the TRAFCON control strategy will be evaluated and compared with the one that can be achieved using the traffic managed currently used by E80.

In the last two months, the TRAFCON team started to develop the planning strategy and proceeded with the construction of the arena.

For the path planning, two main situations have been considered: the startup and the online phase. At startup, where some waiting time is tolerated, an efficient but time consuming multi-robot planning strategy exploiting the traffic measure has been developed. For the online phase, where waiting time  is not be tolerated, a sub-optimal but computationally efficient planning strategy is being developed.

The localization system has been installed in the arena and some experiments have been run for testing it.

In order to implement an effective (re-)planning strategy for the AGVs, in the last two months a measure of efficiency has been implemented. Given a set of paths for the vehicles, this measure evaluates the efficiency both in terms of the traffic that can be generated and in terms of the lengths of the paths: the lower the measure the better the paths. The planning strategy that will be developed will look for set of paths minimizing this measure.

In parallel,  an area for hosting the arena, a small scale automatic warehouse where to carry on the experimental validation in the last part of TRAFCON, has been individuated. The area is 400 m2 wide and the preliminary design and construction of the arena have started.

The last two months of Task 1 have been devoted to the development of a coordination strategy for efficiently solving the coordination problem in a realistic setting. The coordination of the AGVs and the constraints due to the control architecture of Elettric80 have been represented using extended coordination diagrams. In this way, the coordination problem has been transformed into a simpler planning problem in the coordination diagram. A scalable incremental planning algorithm that efficiently solves the coordination among the AGVs has been developed. The validity of the proposed strategy has been shown through simulations over a real plant and comparative analysis with the strategy currently used in Elettric80.