A Multiobjective Path-Planning Algorithm With Time Windows for Asset Routing in a Dynamic Weather-Impacted Environment

• Published in: IEEE transactions on systems, man, and cybernetics. Systems man, and cybernetics. Systems, 2017-12, Vol.47 (12), p.3256-3271
• Main Authors: Sidoti, David, Avvari, Gopi Vinod, Mishra, Manisha, Lingyi Zhang, Nadella, Bala Kishore, Peak, James E., Hansen, James A., Pattipati, Krishna R.
• Format: Article
• Language: English
• Subjects: Algorithms; Approximate dynamic programming; Drills; Dynamic programming; Electric power generation; Environmental impact; Heuristic algorithms; label setting; Marine vehicles; Meteorology; Multiple objective analysis; Nodes; Oceanography; Pareto optimal; Pareto optimization; Route planning; Schedules; ship routing; Shortest path problem; shortest path problem with time windows; Shortest-path problems; uncertainty; Unmanned aerial vehicles; Unmanned helicopters; Vehicle routing; Waypoints; Weather; Windows (intervals)
• ISSN: 2168-2216; 2168-2232
• DOI: 10.1109/TSMC.2016.2573271

This paper presents a mixed-initiative tool for multiobjective planning and asset routing (TMPLAR) in dynamic and uncertain environments. TMPLAR is built upon multiobjective dynamic programming algorithms to route assets in a timely fashion, while considering fuel efficiency, voyage time, distance, and adherence to real world constraints (asset vehicle limits, navigator-specified deadlines, etc.). TMPLAR has the potential to be applied in a variety of contexts, including ship, helicopter, or unmanned aerial vehicle routing. The tool provides recommended schedules, consisting of waypoints, associated arrival and departure times, asset speed and bearing, that are optimized with respect to several objectives. The ship navigation is exacerbated by the need to address multiple conflicting objectives, spatial and temporal uncertainty associated with the weather, multiple constraints on asset operation, and the added capability of waiting at a waypoint with the intent to avoid bad weather, conduct opportunistic training drills, or both. The key algorithmic contribution is a multiobjective shortest path algorithm for networks with stochastic nonconvex edge costs and the following problem features: 1) time windows on nodes; 2) ability to choose vessel speed to next node subject to (minimum and/or maximum) speed constraints; 3) ability to select the power plant configuration at each node; and 4) ability to wait at a node. The algorithm is demonstrated on six real world routing scenarios by comparing its performance against an existing operational routing algorithm.