During my second week I worked more on making an Expansive-Spaces Tree (EST) strategy. I started with reading important sections of chapter 7 from Principles of Motion planning: Theory, Algorithm and implementation. With help, I got to understand the characteristics of sampling-based planners, and the algorithm for roadmap construction and solving queries for the BasicPRM strategy.

Next, I walked through constructing a primitive EST algorithm, which involved growing and merging two trees from the start and goal positions. With this lesson, I got more hands on experience using the PMPL abstractions such as the local planner and distance metric in coding a Motion Planning Strategy (MPStrategy). I also read more on the EST algorithm in chapter 7 of the Principles of Motion planning to help me connect the dots for some parts of the algorithm I already walked through. Then I stared working on optimizing the EST strategy I already implemented.

I looked at the files generated by the EST algorithm and noticed that the samples were generated in random directions. This sparked my first idea to make the tree grow more “in the direction” of the target position (start or goal, depending on the tree in question). I looked over some documentation in the PMPL library to figure out functions already implemented that would help. With help from a Phd student, I was able to get each new configuration to be generated with a closer distance to the target position than the previous. This method helped reduce random generation in areas of little interest in the work space.

I worked through a second idea of generating new configurations with a minimum distance from its “parent” configuration. This method was to avoid clustering of samples around a specific area. Next I met with my mentor and those I work with to discuss my specific topic of research for the summer, which turned out to be Ligand Binding (more on this later). I was walked through explanations what my task is and why this is such an important topic of research, and then I began working on motion planning algorithms related to the project.

After settling in, we began a crash course on Motion Planning which was our main focus of research through the summer. We did some reading on Probabilistic RoadMaps, attended two Phd defenses, ran some experiments with the Parasol Motion Planning Library (PMPL) to test different sampling methods, and had a discussion session to make sure we were all on track.