Non-trivial path planning requires some input knowledge about the environment. Both a priori knowledge, e.g. about obstacles and bathymetry, and data gathered during the mission can be useful. This geographical information can be presented as map layers and typically includes parameters like (Midtgaard et al., 2006):
• Water depth (a priori; updated during the mission)
• Obstacle locations (and possible classifications)
• Forbidden/dangerous zones
• Reliability/age/accuracy of information
• Seabottom type
Traditionally, significant amount of work has been spent on developing memory-efficient representations of maps, such as the quadtree, TIN models and spatial decomposition (Tan et al., 2004). While these methods may reduce memory requirements by several orders of magnitude without loss of information, the time to access information can increase dramatically. Current start-of-the-art computer systems have enough memory storage to make a uniform grid representation a reasonable choice.
For a deliberate architecture, dependent on a high level of planning, the required grid spacing of the map is set in a large part by the environment. If the environment is simple with a sparse obstacle density, only a coarse map is required. However, a complex environment (e.g. a harbour) requires a fine-gridded map. A reactive architecture may not need a map at all, or only a very coarse one, regardless of the environment. A hybrid architecture, with a top deliberate layer and a bottom reactive layer, only needs a plan for the deliberate layer. Thus, it only requires a coarse gridding on the map in any environment. How coarse is dependent both on the mission an on how much of the control is performed by the deliberate and how much is performed by the reactive layer.
Note that the product of the mission, e.g. a bathymetric map, will typically be stored in another, high-resolution format that need not be kept in memory.
Was this article helpful?