We consider multi robot applications, where a human operator monitors and supervise the team to pursue complex objectives in complex environments. Robots, specially at field sites, are often subject to unexpected events that can not be managed without the intervention of the operator(s). For example, in an environmental monitoring application, robots might face extreme environmental events (e.g. water currents) or moving obstacles (e.g. animal approaching the robots). In such scenarios, the operator often needs to interrupt the activities of individual team members to deal with particular situations. This work focuses on human-multi-robot-interaction in these casts. A widely used approach to monitor and supervise robotic teams are team plans, which allow an operator to interact via high level objectives and use automation to work out the details. The first problem we address in this context, is how human interrupts (i.e. change of action due to unexpected events) can be handled within a robotic team. Typically, after such interrupts, the operator would need to restart the team plan to ensure its success. This causes delays and imposes extra load on the operator. We address this problem by presenting an approach to encoding how interrupts can be smoothly handled within a team plan. Building on a team plan formalism that uses Colored Petri Nets, we describe a mechanism that allows a range of interrupts to be handled smoothly, allowing the team to effectively continue with its task after the operator intervention. We validate the approach with an application of robotic water monitoring. Our experiments show that the use of our interrupt mechanism decreases the time to complete the plan (up to 48% reduction) and decreases the operator load (up to 80% reduction in number of user actions). Moreover, we performed experiments with real robotic platforms to validate the applicability of our mechanism in the actual deployment of robotic watercraft. The second problem we address is how to handle intervention requests from robots to the operator. In this case, we consider autonomous robotic platforms that are able to identify their situation and ask for the intervention of the operator by sending a request. However, large teams can easily overwhelm the operator with several requests, hence hindering the team performance. As a consequence, team members will have to wait for the operator attention, and the operator becomes a bottleneck for the system. Our contribution in this context is to make the robots learn cooperative strategies to best utilize the operator's time and decrease the idle time of the robotic system. In particular, we consider a queuing model (a.k.a balking queue), where robots decide whether or not to join the queue. Such decisions are computed by considering dynamic features of the system (e.g. the severity of the request, number of requests, etc.). We examine several decision making solutions for computing these cooperative strategies, where our goal is to find a trade-off between lower idle time by joining the queue and fewer failures due to the risk of not joining the queue. We validate the proposed approaches in a simulation robotic water monitoring application. The obtained results show the effectiveness of our proposed models in comparison to the queue without balking, when considering team reward and total idle time.

Modeling Supervisory Control in Multi Robot Applications

Mohammadi Raeissi, Masoume
2018-01-01

Abstract

We consider multi robot applications, where a human operator monitors and supervise the team to pursue complex objectives in complex environments. Robots, specially at field sites, are often subject to unexpected events that can not be managed without the intervention of the operator(s). For example, in an environmental monitoring application, robots might face extreme environmental events (e.g. water currents) or moving obstacles (e.g. animal approaching the robots). In such scenarios, the operator often needs to interrupt the activities of individual team members to deal with particular situations. This work focuses on human-multi-robot-interaction in these casts. A widely used approach to monitor and supervise robotic teams are team plans, which allow an operator to interact via high level objectives and use automation to work out the details. The first problem we address in this context, is how human interrupts (i.e. change of action due to unexpected events) can be handled within a robotic team. Typically, after such interrupts, the operator would need to restart the team plan to ensure its success. This causes delays and imposes extra load on the operator. We address this problem by presenting an approach to encoding how interrupts can be smoothly handled within a team plan. Building on a team plan formalism that uses Colored Petri Nets, we describe a mechanism that allows a range of interrupts to be handled smoothly, allowing the team to effectively continue with its task after the operator intervention. We validate the approach with an application of robotic water monitoring. Our experiments show that the use of our interrupt mechanism decreases the time to complete the plan (up to 48% reduction) and decreases the operator load (up to 80% reduction in number of user actions). Moreover, we performed experiments with real robotic platforms to validate the applicability of our mechanism in the actual deployment of robotic watercraft. The second problem we address is how to handle intervention requests from robots to the operator. In this case, we consider autonomous robotic platforms that are able to identify their situation and ask for the intervention of the operator by sending a request. However, large teams can easily overwhelm the operator with several requests, hence hindering the team performance. As a consequence, team members will have to wait for the operator attention, and the operator becomes a bottleneck for the system. Our contribution in this context is to make the robots learn cooperative strategies to best utilize the operator's time and decrease the idle time of the robotic system. In particular, we consider a queuing model (a.k.a balking queue), where robots decide whether or not to join the queue. Such decisions are computed by considering dynamic features of the system (e.g. the severity of the request, number of requests, etc.). We examine several decision making solutions for computing these cooperative strategies, where our goal is to find a trade-off between lower idle time by joining the queue and fewer failures due to the risk of not joining the queue. We validate the proposed approaches in a simulation robotic water monitoring application. The obtained results show the effectiveness of our proposed models in comparison to the queue without balking, when considering team reward and total idle time.
2018
Human Multi Robot Interaction, Multi Robot Systems, Reinforcement Learning, Coloured Petri Net, Smart Water Monitoring
File in questo prodotto:
File Dimensione Formato  
main.pdf

accesso aperto

Tipologia: Documento in Pre-print
Licenza: Accesso ristretto
Dimensione 7.83 MB
Formato Adobe PDF
7.83 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11562/979178
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact