Chandarana_cmu_0041E_10468.pdf (12.72 MB)
Human-in-the-loop Mission Planning and Monitoring for Robot Swarms
Robot swarms are large multi-robot systems that use simple, local control laws to produce global emergent behaviors. They are able to self-organize
and coordinate without the use of a centralized mechanism to accomplish tasks otherwise unachievable by a single individual (e.g., in-situ correlative atmospheric data collection). Due to their use of information obtained only from their direct neighbors, these systems are robust to individual robot failures and insertions or removals of swarm members. As a result, robot swarms are scalable.
Their inherent scalability and robustness makes robot swarms suitable for many applications such as search and rescue and surveillance. The work in this thesis focuses on applications known as Swarm Search and Service (SSS) Missions. In SSS missions, which naturally arise from foraging tasks such as search and rescue, the swarm is required to simultaneously search an area while servicing jobs as they are encountered. Jobs must be immediately serviced and can be one of several different job types
– each requiring a different service time and number of vehicles to complete its service successfully. After jobs are serviced, vehicles are returned to the swarm and become available for reallocation. As part of SSS mission planning, human operators must determine the number of vehicles
needed to achieve this balance. The complexities associated with balancing vehicle allocation to multiple as yet unknown tasks with returning vehicles makes this extremely difficult for humans. Previous work assumes
that all system jobs are known ahead of time or that vehicles move independently of each other in a multi-agent framework. This thesis explores the topic of human-in-the-loop mission planning and monitoring for SSS missions. Natural language-based interfaces are designed for intuitive mission definition. Two models are developed to predict the performance of the swarm: the Queuing Model and the Hybrid Model. The Queuing Model is able to predict the performance of the swarm for missions where the swarm movement is constrained (e.g., urban) and the coverage rate of the swarm remains constant while the Hybrid Model builds upon principles in the Queuing Model to handle additional open environments scenarios where the coverage rate dynamically changes with the size of the swarm. These models, when given to human operators, act as a planning tool aid. Operators can rapidly compare system performance across different configurations, leading to more effective mission plans and improved performance. In addition, the Hybrid Model is able to aid operators in maintaining an accurate, real-time situational awareness
of the mission, thereby allowing operators to determine how well the mission is going and if/what errors are occurring. Lastly, to effectively carry out SSS missions, this thesis presents a decentralized method for breaking off robots to reach multiple job sites and rejoining them with the swarm
once service is completed.
and coordinate without the use of a centralized mechanism to accomplish tasks otherwise unachievable by a single individual (e.g., in-situ correlative atmospheric data collection). Due to their use of information obtained only from their direct neighbors, these systems are robust to individual robot failures and insertions or removals of swarm members. As a result, robot swarms are scalable.
Their inherent scalability and robustness makes robot swarms suitable for many applications such as search and rescue and surveillance. The work in this thesis focuses on applications known as Swarm Search and Service (SSS) Missions. In SSS missions, which naturally arise from foraging tasks such as search and rescue, the swarm is required to simultaneously search an area while servicing jobs as they are encountered. Jobs must be immediately serviced and can be one of several different job types
– each requiring a different service time and number of vehicles to complete its service successfully. After jobs are serviced, vehicles are returned to the swarm and become available for reallocation. As part of SSS mission planning, human operators must determine the number of vehicles
needed to achieve this balance. The complexities associated with balancing vehicle allocation to multiple as yet unknown tasks with returning vehicles makes this extremely difficult for humans. Previous work assumes
that all system jobs are known ahead of time or that vehicles move independently of each other in a multi-agent framework. This thesis explores the topic of human-in-the-loop mission planning and monitoring for SSS missions. Natural language-based interfaces are designed for intuitive mission definition. Two models are developed to predict the performance of the swarm: the Queuing Model and the Hybrid Model. The Queuing Model is able to predict the performance of the swarm for missions where the swarm movement is constrained (e.g., urban) and the coverage rate of the swarm remains constant while the Hybrid Model builds upon principles in the Queuing Model to handle additional open environments scenarios where the coverage rate dynamically changes with the size of the swarm. These models, when given to human operators, act as a planning tool aid. Operators can rapidly compare system performance across different configurations, leading to more effective mission plans and improved performance. In addition, the Hybrid Model is able to aid operators in maintaining an accurate, real-time situational awareness
of the mission, thereby allowing operators to determine how well the mission is going and if/what errors are occurring. Lastly, to effectively carry out SSS missions, this thesis presents a decentralized method for breaking off robots to reach multiple job sites and rejoining them with the swarm
once service is completed.
History
Date
2019-09-27Degree Type
- Dissertation
Department
- Mechanical Engineering
Degree Name
- Doctor of Philosophy (PhD)