Information-driven Trajectory Planning for Multi-Agent Target Tracking under Uncertainty

Abstract

Uncertainty from the target’s motion models can grow significantly when the sensor measurements are unavailable. In this regard, trajectory planning algorithms can improve the target tracking performance by choosing the optimal sensor trajectory that allows sensors to obtain the most informative measurement and minimize the uncertainty. Formulating and evaluating the uncertainty is important in target tracking to assess the system performance and to make further decisions. Several information-theoretic functions are proven to be effective to assess the information value of sensor measurements a posteriori and to estimate the value of future sensor measurements a priori [2]. In order to estimate the future information value, a probabilistic sensor measurement model and a motion model can be designed and used. The sensor measurement model can capture the relationship between the target state and sensor measurements [3], and the motion model can capture the target dynamics. Both models can be either designed from first principles or learned from training data. The expected information gain is then computed by taking the expectation of the information value with respect to all possible future measurements based on these probabilistic models. Therefore, the information-driven planning approach can improve target tracking by computing the optimal sensor trajectory by assessing the uncertainty and predicting the utility of future measurements based on sensor and motion models.

Document Details

Document Type
DoD Grant Award
Publication Date
Jan 18, 2023
Source ID
FA86512310003

Entities

People

  • Jaejong Shin

Organizations

  • Air Force Research Laboratory
  • United States Air Force
  • University of Florida

Tags

Fields of Study

  • Computer science

Readers

  • Adaptive Control and Estimation with Uncertainty in Dynamic Systems.
  • Computational Modeling and Simulation
  • Sensor Fusion and Tracking Systems.