Adaptive autonomy in Unmanned Ground Vehicles

From Master Projects
Jump to: navigation, search

has title::Adaptive Autonomy in Unmanned Ground Vehicles
status: finished
Master: project within::Computational Intelligence and Selforganisation
Student name: student name::Armon Toubman
Start start date:=2011/09/12
End end date:=2012/03/12
Supervisor: Mark Hoogendoorn
Second supervisor: Peter-Paul van Maanen
Company: has company::TNO
Thesis: has thesis::Media:Thesis.pdf
Poster: has poster::Media:PosterAAinUGVs.pdf

Signature supervisor



The failure of automation in unexpected situations remains a problem in the field of robotics. With robots lacking recovery mechanisms for all conceivable situations, human supervisors have to monitor robots for failure during operation and take over control when necessary. Adaptive autonomy, or the ability of the robots to change their level of autonomy depending on their situation, offers the prospect of a decreased workload for the supervisor while increasing the performance of the robots. This thesis describes a support model which provides adaptive autonomy using trust models. The support model calculates the trust it has in each robot and in the supervisor, while it monitors their performance and situations. Based on this trust, the support model selects a robot which the supervisor should take control of to increase team performance. The support model was tested with a simulated bomb disposal task. A human supervisor monitored four robots disarming bombs in an unknown environment. The support model provided three types of support. The first type was the presentation of the trust the support model had in each robot to the supervisor. The second type was the presentation of the support model's selection to the supervisor. As the third type, the support model was given complete control over the autonomy of the robots. The performance of the human-robot teams was measured in trials with each support type and compared to the performance in trials where no support was given to the supervisor. The results indicated that the human-robot team performance increased when the support model was given complete control over the autonomy of the robots, compared to the performance when the supervisor had to decide on the autonomy of the robots.