Enhanced Directional Self-Assembly based on Active Recruitment and Guidance

Enhanced Directional Self-Assembly Based on Active Recruitment and Guidance

Nithin Mathews, Anders Lyhne Christensen, Rehan O'Grady, Philippe Rétornaz, Michael Bonani, Francesco Mondada, and Marco Dorigo

Abstract

We introduce enhanced directional self-assembly (EDSA) -- a novel mechanism for morphology growth through the creation of directed connections in a self-assembling multirobot system. In our approach, a robot inviting a physical connection actively recruits the best located neighboring robot and guides the recruit to the location on its chassis where the connection is required. The proposed mechanism relies on local, high-speed communication between connection inviting robots and their recruits. Communication is based on a hybrid technology that combines radio and infrared to provide local relative positioning information when messages are transmitted between adjacent robots. Experiments with real robotic hardware show that EDSA is precise (misalignment of only 1.2° on average), robust (100% success rate for the experiments in this study) and fast (16.1 seconds on average from a distance of 80 cm). We show how the speed and precision of the new approach enable adaptive recruitment and connection in dynamic environments, a high degree of parallelism, and growth of a moving morphology.

Experimental Setup

We placed an extending robot with an open extension slot to its rear (i.e., at 180°) in the center of a circle of 80 cm radius. We placed a free robot at 12 equally separated starting positions on the circle. For each starting position, we considered 8 starting orientations of the free robot. We let the robots execute EDSA for each combination of starting position and starting orientation, which resulted in 96 independent trials. Using an overhead camera, we took images of the robots after a connection had occurred and recorded both the angular precision (we used an open source tool named ImageJ: http://rsbweb.nih.gov/ij/) and recorded the elapsed time for each trial.

In this particular experimental run, the free robot is placed at position 2 (i.e., at 30° w.r.t the extending robot in the center) with orientation 1 (i.e., 0°). The misaligment between the extending robot and the recruit amounts to -0.5° after forming the connection.

Features of EDSA

Adaptive Recruitment. The benefits of EDSA are reduced if the system cannot adapt to changing conditions, for example, the introduction of new free robots, or the malfunction of a recruited robot. Adaptivity in our system comes from the fact that an inviting robot executes the recruitment process once every control step. This gives a connection inviting robot the ability to react to dynamic situations. Furthermore, the recruitment process guarantees an optimal resource allocation; only one neighboring robot is recruited per extension point while all other neighboring robots are left to continue with other tasks.

The extending robot (in the center) opens an extension point at 90°. Initially, the only available free robot is recruited and guided. When a second free robot is introduced at an angle that is smaller with respect to the extension point, the extending robot adapts to the situation and recruits the closer free robot and ignores the initially recruited robot. The initially recruited robot leaves the self-assembly process and becomes available for other tasks.

Enhanced Parallelism. Previous morphology growth approaches have allowed growth in parallel by different extending robots at different parts of the morphology. Often, however, a single robot needs to extend the morphology in more than one direction. Previously, multiple connections to a single extending robot could only happen one after the other. In contrast, our new approach of EDSA supports simultaneous recruitment and guidance of multiple robots. Therefore, the more connections are required (to generate a more sophisticated robot morphology), the more significant the difference in self-assembly time will become between EDSA and the state-of-the-art mechanism.

The extending robot (in the center) opens four extension points at 45°, 135°, 225°, and at 315°. Four free robots are placed around the extending robot. The extending robot recruits the best situated free robot for each extension point and guides each of them to their respective extension point. All four robots are connected within 16 s to form a star morphology.

Parallel growth can also happen on two separate morphologies located close to one another. EDSA allows two neighboring extending robots (i.e., in the same communication range) to extend their morphologies without interfering with each others growth processes.

A chain (top-right) and a star morphology (bottom-left) are under formation in parallel. The extending robot in the chain morphology seeks for a connection at its 180°. A further extending robot is in the center of the star morphology and seeks for a connection at its 45°. The extending robots share the available resources (i.e., the two free robots) without hindering each other and complete both morphologies in parallel.

Morphology Growth in Motion. A real-world morphology growth system may need morphologies to be created while the morphology is in motion. This might be to save time, or because the morphology is already involved in task execution while growth is occurring. Previous systems however, have assumed that the morphology is stationary while it is growing. For the first time, we show that morphology growth can take place while a morphology is moving.

The morphology extending robot is the third link of a chain morphology that is already in motion towards the camera. The extension point is at the rear (i.e., at 180°) of the morphology extending robot. The recruit first aligns to the extension slot, and then approaches the morphology by driving at a velocity that is higher than that of the morphology.