The objective of this research is to develop adaptive coordination and control methods for autonomous sensor networks employed for search and rescue operations, humanitarian demining, and ambient monitoring. The approach is to develop adaptive dynamic programming algorithms for this new class of hybrid systems, to optimize their performance over time by coordinating and implementing future sensor actions intelligently, based on prior knowledge and information obtained by the sensors online.
Intellectual merit This research develops novel hybrid adaptive dynamic programming theory and algorithms for robotic sensor networks. Online learning and adaptive control are crucial to robotic sensor networks because they are, by necessity, deployed in highly unstructured and uncertain environments, with little or no prior information about the obstacles or the targets. This research will merge recent interdisciplinary developments in computer science, robotics, and engineering to develop a novel system-theoretic approach that integrates hybrid control with computational geometry algorithms, and optimizes objective functions derived using information and probability theories.
Broader impact By optimizing the performance of emerging sensor technologies and increasing their level of autonomy, this research will enable them to remove or assist humans in carrying out dangerous yet vital missions, such as humanitarian demining, and rescue efforts following hurricanes, fires, or avalanches. Research results will be disseminated and demonstrated on real systems through existing collaborations with the industry and international institutions. Students will be recruited from underrepresented minorities at both institutions, Duke and UNM, to participate in stimulating education activities, including robotic games, computer game competitions, and environmental research.
As part of this grant, we have developed a transformative and innovative Adaptive Dynamic Programming framework for coordinating heterogeneous mobile sensor networks. Here we briefly discuss a few examples of our achievements. At the University of New Mexico (UNM), we have developed a heterogeneous multi-vehicle test bed (aerial and ground vehicles) to verify and validate the methodologies and computational tools created in this project. The multi-vehicle test bed can be seen as networked hybrid systems where high-level decisions are implemented by adaptive nonlinear controllers. We have theoretically characterized the performance of heterogeneous sensor network for prioritized sensing. An approach that makes use of a heterogeneous team of agents has several advantages when cost, integration of capabilities,or possible large search areas need to be investigated. A heterogeneous team allows for the robots to become specialized in their abilities and therefore accomplish sub-goals more efficiently which in turn makes the overall mission more efficient. The problem of tracking and monitoring moving targets using mobile sensor agents is relevant to a variety of applications, including monitoring of endangered species, civilian security, and surveillance. We have shown that our information potential approach for tracking and surveilling multiple moving targets using mobile sensor agents results in a new approach for computing the motion plans and control inputs of a dynamic sensor agent, based on the feedback obtained from a modified particle filter used for tracking multiple moving targets in a region of interest. A modified particle filter was developed that implemented a sampling mechanism based on supporting intervals of normal probability density functions (PDFs). The method accounts for the latest sensor measurements by adapting a mixture representation of the target probability density functions. The target motion is modeled as a semi-Markov jump process, such that the PDFs of the Markov parameters, can be updated based on real-time sensor measurements by a centralized processing unit. The proposed new information potential method computed an artificial potential function based on the output of the modified particle filter. Using this artificial potential, the sensors compute feedback control inputs that allow them to track and monitor a maneuvering target over time, using a bounded field of view (FOV). We have developed algorithms for agile transportation of suspended loads using aerial robots. This problem has great significance in many UAV applications. We implemented reinforcement learning techniques to generate swing-free trajectories. We have also investigated a coverage control problem where a team of mobile sensors covers a chemical concentration defined by a distributed density function. We showed the Lyapunov stability of an adaptive and decentralized version of the coverage control. This new coverage approach assumed nonholonomic sensors that synchronized themselves through a binary consensus protocol. In addition to theoretical contributions, we verified our proposed algorithms through complex experiments performed both indoors and outdoors using our test bed of aerial and ground robots. Our results have confirmed our hypothesis that heterogeneous systems with different capabilities and sensing modalities can perform talks that would be impossible to carry out using homogeneous agents. On the educational and outreach side, we have had the opportunity of training students throughout the course of this project. We developed and taught a course that covered different aspects of the proposed work "Autonomous Mobile Robots". We furthermore have advised several students including Hispanic and female students to work on coordination of dynamic sensor networks towards their PhD degrees. We have also hosted many K-12 visitors in our labs, with a particular emphasis on Hispanic students. We have developed workshops and small robotic hands-on games for some of the local Hispanic high school students on our campus. We have also trained our own graduate students to take part in working with the incoming high school students to help them develop the hands-on projects. Our overall goal has been to foster the interest of minority and Hispanic students in STEM fields.
