DCER 2014 - Workshop: Distributed Control and Estimation for Robotic Vehicle Networks

Applications for autonomous multi-vehicle networks have grown significantly in recent years, and have stimulated research on distributed strategies for optimal/robust cooperative autonomy in multi-vehicle systems. Ideally, distributed approaches not only perform as well as centralized methods, but also lead to better scalability, naturally parallelized computation, and resilience to communication loss and hardware failures. In practice, it is usually convenient to assume that distributed control and distributed estimation problems can be treated separately. While state-of-the-art techniques for distributed planning (e.g. graph-based trajectory generation, consensus-/graph-based task allocation) and perception (e.g. multi-robot SLAM/SAM, Bayesian/consensus sensor fusion for cooperative tracking) can be combined with good results, the assumed “separation principle” is heuristic and leaves open many questions: how should off-the-shelf solutions for different parts of the same problem be jointly selected or modified to work best together, and what guarantees (if any) are there for optimal/robust behavior? Alternative integrated approaches have also emerged for multi-vehicle systems (e.g. distributed optimization, model predictive control, reinforcement learning), which formally capture and exploit subtle yet important dynamic linkages between the control and estimation problems. However, these approaches raise their own questions: are the assumptions/approximations required for analytical and computational tractability reasonable for general applications, and how can state-of-the-art planning/perception methods for individual mobile robots be leveraged?
This workshop will bring together control/planning and estimation/perception specialists from the robotics and controls communities who are interested in autonomous multi-vehicle networks to: (i) discuss these and other related research questions; (ii) promote new ideas for unifying distributed control and estimation, while improving awareness of state-of-the-art techniques; and (iii) foster interactions for developing theoretical ideas and practical applications.
In addition to attending invited talks by top experts in the field and interacting with them via Q&A panel discussions, workshop participants will have the opportunity to submit extended abstracts for select presentation at the workshop poster spotlight talks and interactive poster session (following single-blind peer review). Selected poster abstract submissions will also be invited to submit their work to a future special issue of IEEE Control Systems Magazine (to be arranged by the workshop organizers).
Some representative topics of interest include (but are not limited to):
optimal/robust multi-robot planning, task assignment, navigation, guidance, and/or control
optimal/robust multi-robot perception, mapping, learning, scene understanding, and/or object search/tracking
control and estimation under ad hoc/constrained/unknown communication topologies
emerging approaches for unified control and estimation (information theoretic methods, learning-based, etc.)
networked algorithmic solutions that combine joint estimation and control
distributed model predictive control
decentralized model-based/Bayesian estimation and learning
distributed optimization for networked robotic vehicle control and estimation
characterization of performance gaps and trade-offs between centralized and distributed algorithms
analysis and algorithms for understanding and coping with uncertainties in networked robotic vehicle systems
novel distributed control and estimation strategies and applications for networks of aerial, space, ground, or aquatic robots
control and estimation of heterogeneous vehicle systems (e.g. mixtures of unmanned ground/air/aquatic robots, marsupial systems, etc.)
distributed control and estimation for long-term/life-long autonomy in robotic vehicle networks