M.Sc. Thesis: Multi-Agent Behavior Tree Synthesis

Multi-agent systems are present in many industrial plants where either co- operative agents help in finishing production faster than a single agent, or where some agents have limited capabilities and must be supported by others (e.g., a robot has a screw driver and another provides the screws). In both cases, agents need to coordinate the task planning and execution. While many approaches exist that achieve such coordination, we plan to use Be- havior Trees which, due to their inherent modularity, offer a flexible and readable graphical models for fast prototyping when it comes to creating new products or variants of existing ones. Additionally, they have a built in execution monitoring mechanism.

Your Task:

The goal of this thesis is the generation of Multi-Agent Behavior Trees (BTs) that allow us to go from a given initial state to a goal state. The BTs will be assigned to different agents available depending on the capabilities of different agents as well as optimization criteria. You will implement and evaluate a planner for assembly tasks. Evaluation will be done in the overcooked environment which will be used as a test bed for high level task planning. You will also compare your results to the previous approaches and the baseline agents mentioned there. Addition- ally, you will solve different challenges faced during the implementation and evaluation of the thesis. In the end, you will hand in a documentation of your results. In the environment, a number of agents are required to cooperate to create a salad consisting of tomatoes and lettuce and other ingredients. The result of your work should be a BT for each agent such that collectively, they can create and deliver the salad.


  • Willingness to learn
  • Solid Pyhton knowledge
  • Experience with multi-agent systems is a plus

Further Information

If you are interested in it or have questions, please feel free to contact Mohamed Behery or Aline Kluge-Wilkes (WZL) via email: