Planning with imperfect information

Abstract

Abstract-We describe an efficient method for planning in environments for which prior maps are plagued with uncertainty. Our approach processes the map to determine key areas whose uncertainty is crucial to the planning task. It then incorporates the uncertainty associated with these areas using the recently developed PAO* algorithm to produce a fast, robust solution to the original planning task.

Topological large-scale off-road navigation and exploration RAVON at the European Land Robot Trial 2008

“…Based on the LTM of E0, it inserted a new direct connection from E0 to P16, starting off in the direction of a well traversable dirt road. 3 Upon initiating the traversal of the new connection, the speculative edge split strategy was invoked repetitively and split the extremely long direct connection iteratively into shorter segments with a length of 20 m each (the used detour node radius d). Using the information obtained from the local short-term memory, the new detour nodes were all placed accurately along the well traversable dirt road.…”

“…[2] uses visual images to extract a global traversability grid map by considering intensity variability and slope direction and performs global path planning using the A*-algorithm. [3] proposes a more robust approach based on prior cost grid maps which finds a path that does not require lengthy detours if grid cells along the route turn out to be intraversable. In [4] , an extension for long range path planning in large grid maps is described which interpolates metrical positions to avoid the 'stepped paths' common to other grid-based path planners and reduces the computational cost of D*-based planning by using variable grid resolutions.…”

RAVON: The robust autonomous vehicle for off-road navigation

“…Over time, they construct a single, large traversability grid map which stores the obtained world information. Longrange path planning is then performed using D*-type algorithms (see (Ferguson and Stentz, 2004) and (Howard et al, 2005) for examples). This approach has been demonstrated to yield good results for semi-rugged terrain such as sparsely vegetated desert or dryland environments.…”