Final Report Summary - MULTI-ROBOT (Multi-Robot Systems)
The overarching research objective of this project is to provide a scientific and practical foundation for the design of multi-robot systems with performance guarantees through an application-centric and interdisciplinary approach. The application-centric approach grounds the theoretical models and algorithms in real-world applications. The chosen application domains are search and logistics, but the results hoped for should transcend these application domains. Both domains require advanced coordination of robots in complex environments with varying degrees of structure and prior knowledge and through isolation of common principles provide a basis for broader results. In addition, real systems in these domains can be realized with current commodity hardware and in collaboration with an emerging logistics & robotics industry. From the overarching objective and the choice of application domains the following concrete objectives are derived:
1. Objective: Comprehensive Multi-Robot Systems for Search
i. Models and algorithm for time optimization for guaranteed search strategies.
ii. Dynamic environments: re-planning and adaptive planning approaches in known and partially known environments; involves the construction of combinatorial representation and their analysis.
iii. Generalized multi-objective optimization (time, resources, planning certainty, control): using algorithms and theoretical results from prior objectives and work this will attempt to generalize the computation of strategies with regard to multiple objectives, formalize their tradeoffs and gain further insights into the properties of search strategies and their computation in environments with complex geometries.
iv. Fully adaptive search: enabling the search dynamic environments with unexpected events and without prior map knowledge; benefits and extends results from the previous objective on dynamic environments and prior work on minimal requirements for guaranteed search without maps.
v. Comprehensive Search system with heterogenous teams, 2.5D and 3D capabilities, human operators and bystanders in dynamic environments: combination of all prior objectives into a comprehensive and extended system; published as open-source; capable of executing searches in unknown environments with multiple robot platforms and in indoor and outdoor environments; full integration of human operator and interfaces; models for dynamic aspects of the environment.
2. Objective: Comprehensive Multi-Robot Systems for Logistics
i. Optimal roadmap-based route optimization and scheduling for multi-robot teams in complex
geometries.
ii. Optimization for dynamic environments and resulting changes in optimal coordination: adaptive optimization for uncertain structural properties of the environment; benefits from results in dynamic environments for search and from adaptive search objectives;
iii. Integration of human operators and models for humans in shared workspaces into the resulting adaptive logistics systems;
iv. Comprehensive logistics system in simulation,
3. Objective: Principles for Multi-Robot Multi-Human Systems
i. Development and application of verification methods for system guarantees with regard to
performance and safety in the presence of human operators and bystanders.
ii. Scalable and efficient integration of human perception into multi-robot systems: detection of targets in search and conflict resolution in logistics.
iii. Utilizing human adaptation to uncertainty and unexpected events in multi-robot systems: crowd-sourcing of perception, error detection and decision-making and integrating these into a scalable and robust system.
iv. Descriptive models of human performance and cognitive models of human behavior for multi-robot system design and verification.
The work performed during the project primarily addresses Objective 1 and 3, with some results regarding Objective 2. The publications [1],[3], [5],[6] relate to Objective 1, [4] relates to Objective 2 and [4] and [7] relate to Objective 3. In addition, source code is made available at https://github.com/Sheffield-Robotics/Multi-Robot-Search which provides the basis for search applications and experiments with real robots. The main results regarding Objective 1 were the analysis of the geometry of environments and their impact on optimization in search problems with some theoretical breakthroughs that allow the efficient computation of search strategies [1]. The main results regarding Objective 2 is a review, taxonomy, and theoretical foundation for the analisys of human-swarm interaction [2]. The joint work from [4] is now being extended to apply these principles in a human-swarm system which has lead to a PhD project [7]. Work regarding Objective 2 commenced in the second period of the project and will be continued beyond this project by a PhD student recruited for this purpose. The student received funding through a University Scholarship Prize from the University of Sheffield. Overall objectives 1 and 3 have been achieved to a satisfactory degree with further results regarding objectives 2 to be expected beyond the termination of the project.
List of relevant publications:
[1] A. Kolling, S. Carpin, A. Kleiner. "Coordinated Search With Multiple Robots Arranged in Line Formations". IEEE Transactions on Robotics, under review.
[2] A. Kolling, P. Walker, N. Chakraborty, K. Sycara, M. Lewis. "Human Interaction with Robot Swarms: A Survey". IEEE Transactions on Human-Machine Systems, 2015, in print.
[3] C. Dornhege, A. Kleiner, A. Hertle, A. Kolling. "Multirobot Coverage Search in Three Dimensions". Journal of Field Robotics, 2015, doi: 10.1002/rob.21573.
[4] J. Chen, M. Gauci, W. Li, A. Kolling, R. Gross. "Occlusion-Based Cooperative Transport with a Swarm of Miniature Mobile Robots". IEEE Transactions on Robotics, vol.PP no.99 1-15, 2015.
[5] H. Qu, A. Kolling, S. M. Veres. "Computing Time-Optimal Clearing Strategies for Pursuit-Evasion Prob- lems with Linear Programming". Towards Autonomous Robotic Systems 2015, 216-228.
[6] H. Qu, A. Kolling, S. M. Veres. "Formulating Robot Pursuit-Evasion Strategies by Model Checking". World Congress of the International Federation of Automatic Control, vol. 19., no. 1. 2014.
[7] G. Kapellmann-Zafra, N. Salomons, A. Kolling, R. Gross. "Human-Robot Swarm Interaction with Limited Situational Awareness", 10th International Conference on Swarm Intelligence, accepted
[8] S. M. Trenkwalder, Y. K. Lopes, A. Kolling, R. Prodan, A. L. Christensen, R. Gross. "OpenSwarm: An Event-Driven Embedded Operating System for Miniature Robots", 2016 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2016)
1. Objective: Comprehensive Multi-Robot Systems for Search
i. Models and algorithm for time optimization for guaranteed search strategies.
ii. Dynamic environments: re-planning and adaptive planning approaches in known and partially known environments; involves the construction of combinatorial representation and their analysis.
iii. Generalized multi-objective optimization (time, resources, planning certainty, control): using algorithms and theoretical results from prior objectives and work this will attempt to generalize the computation of strategies with regard to multiple objectives, formalize their tradeoffs and gain further insights into the properties of search strategies and their computation in environments with complex geometries.
iv. Fully adaptive search: enabling the search dynamic environments with unexpected events and without prior map knowledge; benefits and extends results from the previous objective on dynamic environments and prior work on minimal requirements for guaranteed search without maps.
v. Comprehensive Search system with heterogenous teams, 2.5D and 3D capabilities, human operators and bystanders in dynamic environments: combination of all prior objectives into a comprehensive and extended system; published as open-source; capable of executing searches in unknown environments with multiple robot platforms and in indoor and outdoor environments; full integration of human operator and interfaces; models for dynamic aspects of the environment.
2. Objective: Comprehensive Multi-Robot Systems for Logistics
i. Optimal roadmap-based route optimization and scheduling for multi-robot teams in complex
geometries.
ii. Optimization for dynamic environments and resulting changes in optimal coordination: adaptive optimization for uncertain structural properties of the environment; benefits from results in dynamic environments for search and from adaptive search objectives;
iii. Integration of human operators and models for humans in shared workspaces into the resulting adaptive logistics systems;
iv. Comprehensive logistics system in simulation,
3. Objective: Principles for Multi-Robot Multi-Human Systems
i. Development and application of verification methods for system guarantees with regard to
performance and safety in the presence of human operators and bystanders.
ii. Scalable and efficient integration of human perception into multi-robot systems: detection of targets in search and conflict resolution in logistics.
iii. Utilizing human adaptation to uncertainty and unexpected events in multi-robot systems: crowd-sourcing of perception, error detection and decision-making and integrating these into a scalable and robust system.
iv. Descriptive models of human performance and cognitive models of human behavior for multi-robot system design and verification.
The work performed during the project primarily addresses Objective 1 and 3, with some results regarding Objective 2. The publications [1],[3], [5],[6] relate to Objective 1, [4] relates to Objective 2 and [4] and [7] relate to Objective 3. In addition, source code is made available at https://github.com/Sheffield-Robotics/Multi-Robot-Search which provides the basis for search applications and experiments with real robots. The main results regarding Objective 1 were the analysis of the geometry of environments and their impact on optimization in search problems with some theoretical breakthroughs that allow the efficient computation of search strategies [1]. The main results regarding Objective 2 is a review, taxonomy, and theoretical foundation for the analisys of human-swarm interaction [2]. The joint work from [4] is now being extended to apply these principles in a human-swarm system which has lead to a PhD project [7]. Work regarding Objective 2 commenced in the second period of the project and will be continued beyond this project by a PhD student recruited for this purpose. The student received funding through a University Scholarship Prize from the University of Sheffield. Overall objectives 1 and 3 have been achieved to a satisfactory degree with further results regarding objectives 2 to be expected beyond the termination of the project.
List of relevant publications:
[1] A. Kolling, S. Carpin, A. Kleiner. "Coordinated Search With Multiple Robots Arranged in Line Formations". IEEE Transactions on Robotics, under review.
[2] A. Kolling, P. Walker, N. Chakraborty, K. Sycara, M. Lewis. "Human Interaction with Robot Swarms: A Survey". IEEE Transactions on Human-Machine Systems, 2015, in print.
[3] C. Dornhege, A. Kleiner, A. Hertle, A. Kolling. "Multirobot Coverage Search in Three Dimensions". Journal of Field Robotics, 2015, doi: 10.1002/rob.21573.
[4] J. Chen, M. Gauci, W. Li, A. Kolling, R. Gross. "Occlusion-Based Cooperative Transport with a Swarm of Miniature Mobile Robots". IEEE Transactions on Robotics, vol.PP no.99 1-15, 2015.
[5] H. Qu, A. Kolling, S. M. Veres. "Computing Time-Optimal Clearing Strategies for Pursuit-Evasion Prob- lems with Linear Programming". Towards Autonomous Robotic Systems 2015, 216-228.
[6] H. Qu, A. Kolling, S. M. Veres. "Formulating Robot Pursuit-Evasion Strategies by Model Checking". World Congress of the International Federation of Automatic Control, vol. 19., no. 1. 2014.
[7] G. Kapellmann-Zafra, N. Salomons, A. Kolling, R. Gross. "Human-Robot Swarm Interaction with Limited Situational Awareness", 10th International Conference on Swarm Intelligence, accepted
[8] S. M. Trenkwalder, Y. K. Lopes, A. Kolling, R. Prodan, A. L. Christensen, R. Gross. "OpenSwarm: An Event-Driven Embedded Operating System for Miniature Robots", 2016 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2016)