University of Pennsylvania Department of Electrical and Systems Engineering ESE SENIOR DESIGN PROJECT - ADVISOR PROJECT SUBMISSION FORM 1. Project Title: Hierarchical Information Routing in Dynamic Multi-Agent Systems 2. Proposer's Name: Daniel E. Koditschek, Omur Arslan E-mail: {kod,omur}@seas.upenn.edu Are you willing and able to serve as advisor for this project? _X_ Yes; ___ No Omur would be the Front Line Mentor with backup from Dan 3. Brief Project Description: Notwithstanding their obvious advantages relative to solitary robots, multi-agent systems introduce new challenges requiring efficient and effective solutions for cooperative behaviors, information aggregation, collective decision making and communication. This project aims to design hierarchical information routing methods for mobile multi-agent systems and to evaluate and quantify their performance in an extensible simulation environment built to study practical engineering applications such as distributed sensor fusion and estimation, anomaly detection, coordinated exploration and mapping or environmental monitoring. As the prime motivating setting, consider the problem of multi-robot exploration and mapping in an unknown environment [1],[5],[6]. The integration of individuals’ perceptual constructs (e.g, landmarks or new objects) into a consistent, global map plays a crucial role in such a coordinated exploration and mapping task. To increase efficiency and promote acquisition of new but accurate information, the integration should weigh likely redundancy against task relevance and prioritize the aggregation of local data in assembling the global view. A common method of information collection and routing in stationary wireless sensor networks uses hierarchical clustering of the sensor nodes based on their proximity [2]. Considering a multi-agent mapping problem as a mobile sensor network task motivates exploring analogous hierarchical routing protocols as the basis for a systematic and efficient communications and perceptual integration methodology in settings of distributed robot exploration. One of the critical challenges for mobile agents is that their clustering hierarchy dynamically changes as the agents move, introducing the need for distributed tracking of the clustering tree. Fortunately, it has been known for decades that the relation between minimum spanning trees and single linkage clustering [3,4] yields one such distributed clustering scheme. Hence, a baseline solution for the coordinated mapping task might use such a distributed method [3], in a greedy manner to track the change in clustering hierarchy of the agents by performing re- clustering as required along their collective motion trajectories. Other hierarchical clustering methods of interest to the project proposers could then be considered and compared. 4. Project Design Objectives: (1) Design and implement a computational engine (either from a fresh start or by adapting one already extant [6],[7],[8]) for simulation study of hierarchically-coordinated multirobot exploration and mapping. The engine should permit ready specification of varied indoor or outdoor spatial environments as well as managing algorithms for driving multiple autonomous agents with varied mobility and perceptual capabilities as exemplified in the recent literature [1],[5],[6]. (2) Implement within this simulation environment an instance of the greedy distributed hierarchical clustering method [3,4] for tracking the time varying cluster hierarchy tree arising from the changing configurations of the agents within their environment. (3) Design with consultation from the project proposers an alternative distributed hierarchical clustering method and an evaluation tool for comparing its (and future alternative schemes’) performance against that of the default greedy method. (4) Design with consultation from the project proposers a sample data integration protocol built over the alternative distributed hierarchical clustering methods (2) and (3) and use it to run comparative simulations of: (i) hierarchical map merging, and (ii) hierarchically-coordinated exploration. (5) Evaluate with consultation from the project proposers the effect of various cluster-head (i.e., representative agent of each cluster) selection methods , on the performance of the data integration protocol in (4). 5. Project Prerequisites: What specific knowledge (e.g. courses or topics) and skills (e.g. programming languages or software packages) will this project require? Please rank order the knowledge and skills you have identified, with the most important at the top of the list. Advanced Programming Skills (familiarity with C/C++, Matlab, OpenCV, ROS) Computer Networks (especially channel access methods) Robot Motion Control and Planning Methods Probability Theory and Bayesian Statistics Simultaneous Localization and Mapping (SLAM) Methods 6. References: [1] Fox, D., Ko, J., Konolige, K., Limketkai, B., Schulz, D., & Stewart, B. "Distributed multirobot exploration and mapping." Proceedings of the IEEE 94.7 (2006): 1325-1339. [2] Akkaya, Kemal, and Mohamed Younis. "A survey on routing protocols for wireless sensor networks." Ad hoc networks 3.3 (2005): 325-349. [3] Gallager, Robert G., Pierre A. Humblet, and Philip M. Spira. "A distributed algorithm for minimum-weight spanning trees." ACM Transactions on Programming Languages and systems (TOPLAS) 5.1 (1983): 66-77. [4] Gower, John C., and G. J. S. Ross. "Minimum spanning trees and single linkage cluster analysis." Applied statistics (1969): 54-64. [5] M. A. Hsieh, A. Cowley, J. F. Keller, L. Chaimowicz, B. Grocholsky, V. Kumar, C. J. Taylor, Y. Endo, R. C. Arkin, B. Jung, D. F. Wolf, G. S. Sukhatme, and D. C. MacKenzie, “Adaptive teams of autonomous aerial and ground robots for situational awareness,” Journal of Field Robotics, vol. 24, no. 11–12, pp. 991–1014, 2007. [6] J. Butzke, K. Daniilidis, A. Kushleyev, D. D. Lee, M. Likhachev, C. Phillips, and M. Phillips, “The University of Pennsylvania MAGIC 2010 multi-robot unmanned vehicle system,” Journal of Field Robotics, vol. 29, no. 5, pp. 745–761, 2012. [7] Koenig, Nathan, and Andrew Howard. "Design and use paradigms for gazebo, an opensource multi-robot simulator." Intelligent Robots and Systems, 2004.(IROS 2004). Proceedings. 2004 IEEE/RSJ International Conference on. Vol. 3. IEEE, 2004. [8] Richard Vaughan. "Massively Multiple Robot Simulations in Stage", Swarm Intelligence 2(2-4):189-208, 2008. Springer.