direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Logo der TU Berlin

Details zu Abschlussarbeiten

Zur Übersicht

Thema



Goal-directed Behaviour-based Planning for Multi-Robot Systems

Aufgabenstellung



Description:

Higher-level control of multi-robot systems requires appropriate decision-making and planning in highly dynamic (uncertain) environments. The different existing concepts of reactive decision-making and goal-directed symbolic planning have advantages and disadvantages in such environments. An advantageous solution can be the fusion in a hybrid approach incorporating the advantages of both concepts.

As part of a Bachelor or Master's Thesis you will further improve or extend an already existing hybrid solution that is part of a larger concept. The existing systems consists of a reactive behaviour-network layer and symbolic planner that uses PDDL.
Depending on the particular interests it is possible to focus on specific aspects or to address a wider ranger of smaller issues as well as to integrate your own ideas. The existing solution is tightly integrated into the Robot Operating System (ROS) and consists of modules written in Python as well as C++.

Supervision (Betreuung) is possible in English or German (Deutsch).

Possible topics
1. Further development of the behaviour-network layer
Improve effect to sensor mapping/assignment
Improve concurrency
Other minor extensions → Developing additional activation function kernels
2. Incorporating reinforcement learning into the behaviour-network layer
3. Incorporating another abstraction level for symbolic planning
4. Integration of Task decomposition (also related to topic 3)
centralised vs. decentralised (auction)
5. Integrating a probability concept
6. Case-study about the realisation of different swarm/self-organisation algorithms
7. Developing a graphical notation (inspiration UML) for simplified modelling, interaction and monitoring of the behaviour system
8. Integrating alternative planners

Requirements:
• Python/C++ depending on the particular topic
• Foundation in AI (e.g. Planning, Decision-Making) and Machine Learning depending on the particular topic
• Experience with ROS and Linux are an advantage
• Experience with Qt for UI related topics

Related Literature:
• Philipp Allgeuer and Sven Behnke, “Hierarchical and State-Based Architectures for Robot Behavior Planning and Control,” in Proceedings of 8th Workshop on Humanoid Soccer Robots, IEEE-RAS Int. Conf. on Humanoid Robots, Atlanta, USA, 2013, http://www.humanoidsoccer.org/ws13/papers/HSR13_Allgeuer_Behavior.pdf.
• Young-Seol Lee and Sung-Bae Cho, “A Hybrid System of Hierarchical Planning of Behaviour Selection Networks for Mobile Robot Control,” International Journal of Advanced Robotic Systems, 2014, 1, doi:10.5772/56088.
• David Jung and Alexander Zelinsky, “An Architecture for Distributed Cooperative Planning in a Behaviour-Based Multi-Robot System,” Robotics and Autonomous Systems, Field and Service Robotics, 26, no. 2–3 (February 28, 1999): 149–74, doi:10.1016/S0921-8890(98)00066-9.


If you are interested in this research field please contact me by mail and include a CV as well as your current transcript of records (Notenübersicht).

Betreuer



Betreuer:  Christopher-Eyk Hrabia
Email:   christopher-eyk.hrabiatu-berlin.de

Bearbeiter



Noch Offen