A Tree-Based Algorithm for Construction Robots

In this paper, we present a tree-based algorithm for construction robots. Inspired by the TERMES project of Harvard University, robots in this domain are required to gather construction blocks from a reservoir and build user-specified structures much larger than themselves. While the robots are of roughly the same size as the blocks, they can scale greater heights by using temporarily constructed ramps in the substructures. In this paper, we consider the problem of minimizing the number of pickup and drop-off operations performed on blocks in order to build user-specified structures. Our polynomial-time algorithm heuristically solves this problem and is based on the idea of performing dynamic programming on a spanning tree in the inner loop and searching for a good tree to do so in the outer loop. Our algorithm performs very well in simulation and scales easily to large problem instances. For planning problems of this nature that are akin to construction domains, we believe that valuable lessons can be learned from comparing the success of our algorithm with the failure of off-the-shelf planning technologies.