ECE 595 (Autonomous Mobile Robots) Project web page
Optimal Coordination of a Team of Mobile Agents for Active Monitoring in Presence of Realistic Communication Channels

Project Description:

In summary, we intend to study the problem of optimal coordination of a team of mobile agents that are tasked to monitor an unknown environment. The agents are equipped with non-ideal sensors whose sensing qualities degrade as the distance to the nodes increases. A fusion center is in charge of fusing all the data from the agents to detect any abnormality in the environment. In order to maximize the probability of detection, the fusion center should navigates the agents to the points that give the best sensing quality (or best coverage). The problem will then be how to design the controller laws to achieve this goal. In case of perfect communication links between the agents and the fusion center, the problem has been already solved and the optimal configurations are given by Voronoi partitions. However in presence of non-ideal communication links, the problem becomes considerably challenging. In order to maintain the connectivity of the agents with the fusion center, the motion coordinator has to consider the full Signal Strength or Signal-to-Noise Ratio (SNR) map of the environment to avoid the regions with poor channel quality. Note that connectivity maintenance in this particular problem is very important as loosing connection with one node means loosing control over a large area of the environment. Consequently, we try to formulate an constraint optimization problem to be solved at the fusion center in real-time in order to navigate the robots to the points that give the best possible sensing quality and coverage while maintaining their connectivity. Unfortunately, the optimization problem in this level needs a reliable SNR map of the environment which is usually not available for practical applications. We then propose a channel learning scheme to be combined with the previous technique. Consequently, the contribution of the work will be two fold:

We first solve the optimal coverage control problem in a communication-aware case with considering the available SNR map of the environment.

We then combine this technique with a channel learning strategy to update the SNR environment as the node move.

Progress Reports:

Methamtical Description of the Problem: In this part we focus on precise mathematical description of the problem. The problem as it is, even for simple communication settings is very challemging. We also try to simplify the problem to be solvable. The theoretical results of this part is available upon request.
Channel Learning Strtegies: In this part we focused on a wireless channel learnning algorithm to be deployed by the robots along their trajectories. The channel estimation is based upon a new algorithm that uses a combination of Kalman filtering and a new moving window dynamic BLUE. The kalman filter is used to dynamically estimate the path-loss parameters and the moving window BLUE is estimating the shadowing part using a fixed number of observations near the current location of the robot. The resulting map gives a local estimate of the SNR variations which will be fed to the motion planner for planning the motion of the robot. So far we have developed the theoretical part including a recursive update rule for estimating the shadowing part. Currently we are working to simulate a simple scenario using the proposed technique.
- The thoretical results for channel estimation will be available upon request.
- The Matlab source code for simulating can be downloaded from here.