For biomedical applications in targeted therapy delivery and interventions, a large swarm of micro-scale particles (“agents”) has to be moved through a maze-like environment (“vascular system”) to a target region (“tumor”). Due to limited on-board capabilities, these agents cannot move autonomously; instead, they are controlled by an external global force that acts uniformly on all particles. In this work, we demonstrate how to use a time-varying magnetic field to gather particles to a desired location. We use reinforcement learning to train networks to efficiently gather particles. Methods to overcome the simulation-to-reality gap are explained, and the trained networks are deployed on a set of mazes and goal locations. The hardware experiments demonstrate fast convergence, and robustness to both sensor and actuation noise. To encourage extensions and to serve as a benchmark for the reinforcement learning community, the code is available at Github.

We illustrate and animate the classic problem of deciding whether a given graph has an Eulerian path. Starting with a collection of instances of increasing difficulty, we present a set of pictorial instructions, and show how they can be used to solve all instances. These IDEA instructions ("A series of nonverbal algorithm assembly instructions") have proven to be both entertaining for experts and enlightening for novices. We (w)rap up with a song and dance to Euler’s original instance.

We motivate, visualize and demonstrate recent work for minimizing the total execution time of a coordinated, parallel motion plan for a swarm of N robots in the absence of obstacles. Under relatively mild assumptions on the separability of robots, the algorithm achieves constant stretch: If all robots want to move at most d units from their respective starting positions, then the total duration of the overall schedule (and hence the distance traveled by each robot) is O(d) steps; this implies constant-factor approximation for the optimization problem. Also mentioned is an NP-hardness result for finding an optimal schedule, even in the case in which robot positions are restricted to a regular grid. On the other hand, we show that for densely packed disks that cannot be well separated, a stretch factor Omega(N^{1/4}) is required in the worst case; we establish an achievable stretch factor of O(N^{1/2}) even in this case. We also sketch geometric difficulties of computing optimal trajectories, even for just two unit disks.