Yoshida’s paper addresses a reconfiguration planning method for locomotion of a homogeneous modular robotic system. The experiment is to verify that the planned locomotion can be realized by hardware. The paper discusses a two-layered planning method for locomotion of a class of regular structures. The planning simulation demonstrates that this approach effectively solves the complicated planning problem. This structure enables the robot to move in three dimensions by changing its configuration. One of the issues posed by this problem is that a vase search space must be explored to examine the interchangeability between two arbitrary module configurations and to avoid collisions between modules in 3D space.
The architecture of the two layer approach consists of a global flow planner and the local motion scheme selector. The global flow planner searches possible module paths and motion order to provide the global cluster movement, called a flow. This rule includes a local reconfiguration motion sequence, called a motion scheme. It is interesting that in this research, there are several methods of generating this cluster motion. The paper chooses a simple motion that sends modules toward the head, or primarily a forward-roll motion on the side of the cluster.
There are a few assumptions made within the research. First, the research assumes that only one motion scheme is allowed at one time and that the flow direction does not self-intersect and runs straight for at least two unit lengths during the cluster flow. By making this assumption, it limits the configurations that the cluster could reconfigure itself to because there are situations where you would want to reconfigure yourself simultaneously from different sides. The other issue is with the assumption that in the planning that one module can lift only one other module, which comes from the limited torque capacity of the hardware. There are situations where the module would need to lift more than one module, especially for more complex shapes depending on the situation. This is a hardware constraint imposes simply by the physics of the cluster and therefore there doesn’t seem to be a method for fixing this issue.
The key to extending this architecture would to extend the rule sets so that there is a wider class of configurations for the objects. Another piece would to enable sensors, or at the very least touch sensors so that the cluster can adapt when it comes into contact with a wall or some other object. These objects also have the potential to adapt to any kind of environment if the modules can move in a fashion such that the cluster can reconfigure itself when give a constrained environment. For instance, the cluster would be able to go navigate through all the debris when searching for survivors due to its dynamic reconfiguration.
Reference:
- Yoshida, Eiichi. “A Self-Reconfigurable Modular Robot : Reconfiguration Planning and Experiments.” <http://www.researchgate.net/publication/228746149_A_self-reconfigurable_modular_robot_Reconfiguration_planning_and_experiments/links/00b4952aff8f15f81d000000.pdf>
