Auftragsbasierte Führung von Mehrrobotersystemen mit quasi-natürlicher Sprache

  • Mission command based control of multi-robot systems with quasi-natural language

Remmersmann, Thomas; Schlick, Christopher (Thesis advisor); Flemisch, Frank Ole (Thesis advisor)

Aachen (2016)
Dissertation / PhD Thesis

Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2016


The ongoing developments in automation and sensor technology make it feasible to perform more and more complex tasks using robots. Currently, attempts are being made to let robots handle increasingly complicated and more dynamically variable tasks. Many of these tasks often cannot be efficiently carried out by a single robot. In such cases, a group of cooperating robots, known as a Multi-Robot System (MRS), is needed. However, the control of an MRS by a single operator is only possible if he is supported by assistant functions.This dissertation describes the conceptualization and development of a prototypical assisting robot command and control system (C2 system) using a quasi-natural language for mission command. Command and Control Lexical Grammar (C2LG) is a standardized grammar, which can be used to command MRSs. This dissertation presents the Multi-Robot Behavior Description and Control Language (MRBDCL). This language is based on C2LG and supports the definition of robot behavior. The concept development describes how aspects of issuing, processing and interpreting commands have been incorporated into the design of MRBDCL. In addition to the consideration of these technical requirements, it was also a goal to construction MRBDCL in a way that feels quite natural, thus making it easy to learn. One main consideration in the implementation was the optimization of command execution during runtime. Optimization of command execution was achieved through the use of genetic algorithms. We examined which mutation and recombination operators are best suited for given problems. Technical validation of the prototype was accomplished through two experiments with real robots. The first experiment demonstrated the working interaction concept based on the C2LG and in mission command based style using a heterogeneous MRS with unmanned ground vehicles (UGVs) and unmanned aerial vehicles (UAVs). In the second, the planning system was validated, showing that reconnaissance of a road network and measurement of the concentration of hazardous materials in an area could each be initiated by a single C2LG task. The details of the tasks were determined by the planning system.Validation of the software ergonomic of the C2 system for an MRS was carried out in a study under controlled conditions. The test persons had to use the C2 system in different simulated scenarios. They were tasked with locating hazmat signs using the robots. When the robots detected a hazmat sign, the operators were to check the accuracy of the detection and carry out corrections when necessary. The results showed that the operators performed higher when using the MRBDCL-based group control compared to a single robot task assignment. The study finishes with the interpretation of eye-tracking and user survey data to identify possible improvements for the prototype.