What exactly is LPA*? I haven't been able to find any easy to understand descriptions from Google, but so far from what I can gather it's just a mix of path caching and a B*-style algorithm? Here's how I imagine it works: - Start an A* search from the start node to the end node - Start a backwards A* search from the end node to the start node - Keep going until the current node in the first search and the current node in the second search match the start node and the end node for a cached path Is this correct?

Im not familiar with LPA* but what youi define sounds like a way to parallelize A*. But it would have to be a collision of the closed lists triggering the end because it would be rare for the two actively tested nodes to be the same one.

The shared data (closed list) would take some interlocking to prevent data consistancy problems .