Apart from reasoning to determine a plan of actions to accomplish a goal, robustly executing high-level plans on mobile robots often results in additional problems. In particular, the robot platform poses additional constraints to the execution, e.g., the robot's gripper needs to be calibrated before it can be used.
In previous work as part of the ConTrAkt project, we have developed an approach to deal with such additional constraints : First, we define the abstract agent behavior with the agent programming language Golog . Additionally, the robot platform and its constraints are modeled with timed automata and the logic t-ESG . The abstract plan is then transformed into a platform-specific plan by applying reachability analysis for timed automata .
In this thesis, the goal is to apply this approach to logistics robots in the context of the RoboCup Logistics League. In particular, the goals of this thesis are:
- Extend and adapt the existing Golog++ implementation of an RCLL agent.
- Develop platform models for the Carologistics Robotinos.
- Formulate platform constraints based on the platform model.
- Integrate the existing plan transformation  in the agent.
What we offer:
In this thesis, you will
- work on real robots with a real-world application in a competitive scenario;
- be part of the Carologistics RoboCup team, current world champion in the RoboCup Logistics League;
- contribute to an open-source project, publicly visible on GitHub.
What we expect:
- be familiar with C++,
- have a basic understanding of logic, ideally with some background in temporal logic.
Additionally, the following qualifications are beneficial:
- You have attended one of our lab courses.
- You have attended the lecture Introduction to Knowledge Representation.
- You have some experience with mobile robots.
- You are familiar with Linux generally and with Fedora Linux specifically.
If you are interested or have further questions, please contact Till Hofmann.
T. Hofmann, V. Mataré, S. Schiffer, A. Ferrein, and G. Lakemeyer, “Constraint-based online transformation of abstract plans into executable robot actions,” in AAAI Spring Symposium: Integrating Representation, Reasoning, Learning, and Execution for Goal Directed Autonomy, 2018. PDF. ↩︎
T. Hofmann and G. Lakemeyer, “A logic for specifying metric temporal constraints for Golog programs,” in Proceedings of the 11th Cognitive Robotics Workshop 2018 (CogRob), Tempe, AZ, USA, 2018. PDF. ↩︎