More and more shopfloors see the integration of fleets of autonomous guided vehicles
(AGVs) to execute transport tasks. Previously, the control of these robot fleets was
predominantly done by central controllers. This severely limits the scalability of such
systems and introduces a single point of failure, to just name two disadvantages. With
the increasing size of typical use cases, a trend toward the decentralization of control
tasks can be observed. Additionally, many factories exhibit a special structure in their
environment, resulting from the modern robot hardware that is available. Precisely, the
system consists of fully connected areas and their connections. So far, this information
is not utilized by typical deadlock handling strategies.

This thesis introduces a hierarchical environment representation that can capture the
described structure of the environment efficiently. On this basis, a hybrid deadlock
prevention approach is proposed that uses multiple, independent components. For
this, several novel strategies are derived which can be used to solve the problem of
positioning idle robots, and their consequences on the characteristics of the system are
explored. The proposed hybrid approach is then compared to a reference algorithm
to evaluate its performance. Simulations are conducted in order to compare the
two methods concerning the most important metrics. It is shown that the developed
deadlock prevention approach outperforms a reference algorithm in terms of the system
throughput, which is the key metric for this use case. This result is achieved even
while subjecting the used robots to lower wear and tear, as well as lower energy usage
compared to the reference algorithm because fewer unnecessary detours are made.