2018/07/31 by Patrick Slade, Slade, Patrick, Zachary N. Sunberg +3 · 1 citation
Engineering · #Advanced Control Systems Optimization #Control Systems and Identification #FOS: Electrical engineering #Fault Detection and Control Systems #Systems and Control (eess.SY) #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.1808.00888
openalex publication_date 2018/07/31 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/01
Real-world autonomous systems operate under uncertainty about both their pose\nand dynamics. Autonomous control systems must simultaneously perform estimation\nand control tasks to maintain robustness to changing dynamics or modeling\nerrors. However, information gathering actions often conflict with optimal\nactions for reaching control objectives, requiring a trade-off between\nexploration and exploitation. The specific problem setting considered here is\nfor discrete-time nonlinear systems, with process noise, input-constraints, and\nparameter uncertainty. This article frames this problem as a Bayes-adaptive\nMarkov decision process and solves it online using Monte Carlo tree search with\nan unscented Kalman filter to account for process noise and parameter\nuncertainty. This method is compared with certainty equivalent model predictive\ncontrol and a tree search method that approximates the QMDP solution, providing\ninsight into when information gathering is useful. Discrete time simulations\ncharacterize performance over a range of process noise and bounds on unknown\nparameters. An offline optimization method is used to select the Monte Carlo\ntree search parameters without hand-tuning. In lieu of recursive feasibility\nguarantees, a probabilistic bounding heuristic is offered that increases the\nprobability of keeping the state within a desired region.\n