This page is under construction. Please be patient!

Like its predecessor, Argos employs a three-pronged approach to navigation: a cost map is generated to represent the environment, a desired waypoint is selected, and finally, the path planner calculates an optimal trajectory to the targeted waypoint.

While the path generation and waypoint selection mechanisms remain largely the same, Argos significantly improves upon Kratos in its path planning stage. As a replacement for Kratos’s simplistic A* path planner, Argos uses the Dynamic A* Lite algorithm. By planning backwards from the waypoint, D* Lite avoids unnecessary searches even as the robot moves, increasing computational speed without sacrificing results.

In addition, Argos also employs field searching in its path planning stage. Rather than restrict the search nodes to cost cell centroids, which would only result in paths that moved neatly from centroid to centroid, and only allow for turns in increments of 45 degrees, field search is considerably more flexible. A traditional path planning algorithm searches for the minimum cost path to s from each of its neighbors s’. Field search also considers paths to s from any point on the boundary between neighboring nodes, and linearly interpolates the depth of novel predecessors to determine cost. This allows for a smoother, more optimal path.

To offset the additional computational cost posed by this more thorough search, Argos optimizes using such methods as a binary heap, cached costs, and compiling with the SSE3 instruction set extensions.

All in all, our Field D* path planner operates at roughly 10 Hz, about the same speed as the less sophisticated A* planner used by Kratos.