The proposed research will integrate control, sensing and planning to develop autonomous systems that can effectively perform search, surveillance, and pursuit. The goals of the research are to develop planning and control algorithms for systems with limited or constrained sensor capabilities, to design motion strategies that take account of geometric, kinematic, and dynamic constraints, and to understand and balance the trade-off between the optimality and complexity of the required algorithms. These goals will be achieved by extending previous work to deal with geometric constraints in 3D environments (i.e., the presence of obstacles in the world), and with and kinematic and dynamic constraints on pursuer motion; designing, formally verifying and implementing a robot architecture for reactive motion planning; and testing the approach using real robot systems.
(2) Broader Significance:
Robotics is poised to have broad impact on society, and the proposed research has the potential provide inroads to solutions for numerous problems that are of interest today. These include disaster relief, search and rescue, scientific studies (e.g., arctic, undersea or planetary exploration). In addition, the basic mathematical tools that will be developed could find applications in communications, economics and other domains in which uncertainty is relevant to decision making.
The specific focus of the proposed research is on the integration of sensing, planning and control in sensor-based robotics. The proposed approach posits that the role of a planning system is to automatically construct feedback control policies along with rules that can be used to switch between these policies based on sensor input. The need for planning and sensing arises due to the complexity and unpredictability of the world in which the system must operate. The use of feedback control as the fundamental building block for planning allows rigorous characterization of system performance, including stability and convergence.
