2021, Weber, Lukas, Dornberger, Rolf, Hanne, Thomas, Abraham, Ajith, Hanne, Thomas, Castillo, Oscar, Gandhi, Niketa, Nogueira Rios, Tatiana, Hong, Tzung-Pei

2021, Applewhite, Timothy, Zhong, Jia, Dornberger, Rolf

Multimodal interaction is an essential prerequisite for affective human-robot engagement. Research on bidirectional, affective multimodal interaction systems investigates systems that recognize a user's affect and generate emotional response based on this user's affect. The presented work investigates a novel bidirectional, affective multimodal interaction system using a social robot and an open-source dialogue system framework, developing a prototype based here on Pepper and Rasa. Compared to special lab robotics systems, the proposed system is more attainable, while incorporating, alongside speech and facial expression, eye gaze as one of the major features to convey emotions as input channels. The system generates and emulates emotional output behaviors based on a user's affect using speech, gestures and emojis. This paper describes the concrete implementation and evaluation of the proposed system. Results of the evaluation show that, although the recognition accuracy of the input channels perform differently well, the system can derive well-defined rule-based emotional output behaviors with a high multimodal accuracy rate in the given test scenarios.

2021, Kempter, Philipp, Schmitz, Martin Peter, Hanne, Thomas, Dornberger, Rolf, Abraham, Ajith, Hanne, Thomas, Castillo, Oscar, Gandhi, Niketa, Nogueira Rios, Tatiane, Hong, Tzung-Pei

2020, Troxler, Dominik, Hanne, Thomas, Dornberger, Rolf

2021, Wild, Simon, Parlar, Soyhan, Hanne, Thomas, Dornberger, Rolf

This paper analyses how the required skills in a job post can be extracted. With an automated extraction of skills from unstructured text, applicants could be more accurately matched and search engines could provide better recommendations. The problem is optimized by classifying the relevant parts of the description with a multinomial naïve Bayes model. The model identifies the section of the unstructured text in which the requirements are stated. Subsequently, a named entity recognition (NER) model extracts the required skills from the classified text. This approach minimizes the false positives since the data which is analyzed is already filtered. The results show that the naïve Bayes model classifies up to 99% of the sections correctly, and the NER model extracts 65% of the skills required for a position. The accuracy of the NER model is not sufficient to be used in production. On the validation set, the performance was insufficient. A more consistent labelling guideline would be needed and more data should be annotated to increase the performance.

2021, Applewhite, Timothy, Kaufmann, Roger, Dornberger, Rolf, Hanne, Thomas

This paper presents an improved controller based on Ant Colony Optimization for the ghost team of Ms. Pac-Man. The controller is an enhanced version of fairGhosts which is based on the ghostAnt framework. Various improvements were implemented in terms of parameters and concepts. Especially for the explorer ants, fairGhosts uses a simplified version, as the exact reasoning for the proposed concepts could not be determined in the ghostAnt framework. In this paper, two new types of ghost teams are proposed after the modifications were conducted: powerGhosts and defensiveGhosts. powerGhosts take into account the power pill aspect for stopping criteria and solution quality of the explorer ants, and defensiveGhosts additionally involve a threshold of Ms. Pac-Man’s distance to the nearest power pill for the hunter ants, so as not to be caught easily. Test results show that the powerGhosts version shows on average 15% better results than the initial fairGhosts setup, while defensiveGhosts performs equal or slightly worse than the initial implementation. It can be concluded that including the power pill aspect in the explorer ants concept shows an improved performance of the ghost team. On the other hand, the concept of distancing ghosts from Ms. Pac-Man when she is near a power pill did not result in any significant improvement.

2020, Kussmann, Simon, Godat, Yannick, Hanne, Thomas, Dornberger, Rolf

The Capacitated Vehicle Routing Problem (CVRP), an extension of the Traveling Salesman Problem with two added constraints, a local depot and a capacity constraint for each vehicle, is solved by a Hybrid Bat Algorithm (HBA). This paper investigates how the standard Bat Algorithm must be extended to become a HBA being able to solve the CVRP. The Hybrid Bat Algorithm is tested and compared to three other optimization algorithms for the CVRP, the Clarke & Wright Savings Algorithm, the Holmes and Parker Algorithm, and the Fisher and Jaikumar Method. It is discussed how the HBA is able to deliver decent solutions of the CVRP.

2021, Arockia, Amala, Lochbrunner, Markus, Hanne, Thomas, Dornberger, Rolf

This paper addresses the Capacitated Vehicle Routing Problem (CVRP) consisting of a single depot and several customers that are supplied with goods by capacitated vehicles from a depot. The main objective of the vehicle routing problem is to minimize the traveled distance of all vehicles. We compare the Tabu Search (TS) and Simulated Annealing (SA) algorithm with different initial solution strategies to solve the CVRP. We run the publicly available solver on a set of benchmark problems comparing above mentioned methods and initial solutions. The results show that TS appears superior for small-sized problems, while SA has an advantage for mid-sized problems. For larger problems the preferability of a methods depends on the available run time with SA appear promising for shorter runtime and TS for longer.

2021, Oller, Heide, Dornberger, Rolf, Hanne, Thomas, Thampi, Sabu M., Krishnan, Sri, Hegde, Rajesh M., Ciuonzo, Domenico, Hanne, Thomas, Kannan R., Jagadeesh

2020, Faust, Oliver, Mehli, Carlo, Hanne, Thomas, Dornberger, Rolf