Distributed sensor resource management and planning

Deepak Khosla, James Guillochon, Ivan Kadar
2007 Signal Processing, Sensor Fusion, and Target Recognition XVI  
The goal of sensor resource management (SRM) is to allocate resources appropriately in order to gain as much information as possible about a system. In our previous paper, we introduced a centralized non-myopic planning algorithm, C-SPLAN, that uses sparse sampling to estimate the value of resource assignments. Sparse sampling is related to Monte Carlo simulation. In the SRM problem we consider, our network of sensors observes a set of tracks; each sensor can be set to operate in one of several
more » ... modes and/or viewing geometries. Each mode incurs a different cost and provides different information about the tracks. Each track has a kinematic state and is of a certain class; the sensors can observe either or both of these, depending on their mode of operation. The goal is to maximize the overall rate of information gain, i.e. rate of improvement in kinematic tracking and classification accuracy of all tracks in the Area of Interest. We compared C-SPLAN's performance on several tracking and target identification problems to that of other algorithms. In this paper we extend our approach to a distributed framework and present the D-SPLAN algorithm. We compare the performance as well as computational and communications costs of C-SPLAN and D-SPLAN as well as near-term planners. INTRODUCTION Sensor Resource Management (SRM) is "the control problem of allocating available sensor resources to obtain the best awareness of the situation." 3 SRM is important in terms of the benefits it provides over non-coordinated sensor operation. By automating the process, it reduces the operator workload. The operator defines the sensor tasking criteria instead of controlling multiple sensors individually by specifying each operation to be performed by each sensor. In an automated SRM system, the operator concentrates on the overall objective while the system works on the details of the sensor operations. Additionally, the feedback within the SRM system allows for faster adaptation to the changing environment. Problems that SRM has to deal with include insufficient sensor resources, highly dynamic environment, varied sensor capabilities/performance, failures and enemy interference, etc. Desired characteristics of a good sensor manager are that it should be goal oriented, adaptive, anticipatory, user friendly, require minimal interaction, account for sensor dissimilarities, perform in near real time, handle adaptive length planning horizons, etc. Most SRM problems can be formulated as belonging to the class of Markov Decision Process (MDP) problems. In a MDP, future states are assumed to be the result of applying actions to the current state, ignoring the total history of the state space. In centralized methods, nodes with complete knowledge of the state space (Full-awareness nodes, or FANs) maintain the current state, and update the state with incoming measurements or the passage of time. This new state can now be used to determine future states. Similarly, in a decentralized system, each node's state is only dependent on that node's prior state, input from the environment, and process noise, just like any other Markovian process. Each node maintains its own individual picture of the state space, but since inter-node communication is imperfect, these state representations are often dramatically different from node to node. Yet we must remain cautious: If communication times are non-zero, measurement information received from another node will have aged by an amount of time equal to the time required to send the measurement from one node to the other. This can be a problem if the state has already been advanced due to the passage of time. For example: Take a state at t 0 , and now advance this state to t 1 . A measurement is now received from a sensor that was made at t 0 . We would want to update the state at t 0 with the new measurement, and then update the state due to the passage of time. Since we do not know when packets will be received, this requires a set of states to be maintained in order to correctly predict future states. This violates our Markov conditions. One solution to this problem is to assume
doi:10.1117/12.719929 fatcat:ps2caessojbebecwmcyzge5u3u