Generation of a Reconfigurable Probabilistic Decision-Making Engine based on Decision Networks: UAV Case Study (Interactive Presentation)

Making decisions under uncertainty is a common challenge in numerous application domains, such as autonomic robotics, finance and medicine. Decision Networks are probabilistic graphical models that propose an extension of Bayesian Networks and can address the problem of Decision-Making under uncertainty. For an embedded version of Decision-Making, the related implementation must be adapted to constraints on resources, performance and power consumption. In this paper, we introduce a high-level tool to design probabilistic Decision-Making engines based on Decision Networks tailored to embedded constraints in terms of performance and energy consumption. This tool integrates high-level transformations and optimizations and produces efficient implementation solutions on a reconfigurable support, with the generation of HLS-Compliant C code. The proposed approach is validated with a simple Decision-Making example for UAV mission planning implemented on the Zynq SoC platform