AAU Student Projects - visit Aalborg University's student projects portal
A master's thesis from Aalborg University
Book cover


A shared control approach for obstacle avoidance in robot supported USAR

Authors

;

Term

4. semester

Education

Publication year

2021

Submitted on

Pages

106

Abstract

As robotics advances, new ways for people and robots to work together emerge. A key area is shared control, where both the human operator and the robot contribute to the task. The level of autonomy can range from full manual control (teleoperation) to full autonomy. A typical example is Urban-Search-and-Rescue with a mobile robot, where the operator has only a high-level map of the building while the robot senses nearby obstacles that must be avoided. In this study, we examine whether different autonomy levels during obstacle avoidance affect task performance. Participants control a TurtleBot3 in the Gazebo simulator of a house, where obstacles are invisible to the operator and must be avoided to complete the task. We test three conditions with increasing robot autonomy: teleoperation, shared control, and full autonomy. The robot can initiate changes in autonomy, but operators can always take back control. Participants receive audio cues when the autonomy level changes. In the (semi-)autonomous modes, we use a potential field method for reliable and efficient obstacle avoidance, which treats obstacles like repelling forces. We measure how quickly participants complete the task and find no significant differences between conditions. For subjective measures, we use the QUEAD and NASA-TLX questionnaires and find only one significant partial result in favor of teleoperation. Overall, we observe a general trend—both in objective and subjective measures—favoring teleoperation. We conclude that participants may feel more comfortable having full control over the robot, but further research is needed to explore additional scenarios, autonomy modes, and ways to transfer and regain control.

Efterhånden som robotteknologien bliver mere avanceret, opstår nye muligheder for at lade mennesker og robotter samarbejde. Et centralt område er delt kontrol, hvor både operatør og robot bidrager til at løse en opgave. Graden af autonomi kan variere fra fuld manuel styring (teleoperation) til fuld autonomi. Et typisk eksempel er Urban-Search-and-Rescue med en mobil robot, hvor operatøren kun har et oversigtskort over bygningen, mens robotten lokalt kan sanse forhindringer, som den skal undgå. I dette studie undersøger vi, om forskellige autonominiveauer under undgåelse af forhindringer påvirker opgavens resultat. Deltagere styrer en TurtleBot3 i en Gazebo-simulator af et hus, hvor forhindringerne er usynlige for operatøren og skal undgås for at gennemføre opgaven. Vi afprøver tre betingelser med stigende grad af robotautonomi: teleoperation, delt kontrol og fuld autonomi. Robotten kan selv tage initiativ til at ændre autonominiveau, men operatøren kan altid overtage styringen igen. Deltagerne får lydfeedback, når autonominiveauet ændres. I de (semi-)autonome tilstande bruger vi en potentialefeltmetode til pålidelig og effektiv forhindringsundgåelse, hvor forhindringer behandles som frastødende kræfter. Vi måler, hvor hurtigt deltagerne gennemfører opgaven, og finder ingen signifikante forskelle mellem betingelserne. For de subjektive målinger bruger vi QUEAD- og NASA-TLX-spørgeskemaerne og finder kun ét signifikant delresultat til fordel for teleoperation. Samlet ser vi dog en generel tendens – både i objektive og subjektive mål – der favoriserer teleoperation. Vi konkluderer, at deltagerne muligvis føler sig mere trygge ved at have fuld kontrol over robotten, men at der er behov for yderligere forskning i flere scenarier, andre autonomitilstande og metoder til at overdrage og genvinde kontrol.

[This apstract has been rewritten with the help of AI based on the project's original abstract]