The main scientific objective of the proposed research is the design, implementation and control of a novel distributed robotic system made of heterogeneous, dynamically connected small autonomous robots in order to form a swarmanoid. The swarmanoid that they intend to build will be comprised of numerous (about 60) autonomous robots of three types: eye-bots, hand-bots, and foot-bots.

Eye-bots are specialized in sensing and analyzing the environment from a high position to provide an overview that foot-bots or hand-bots cannot have. They are based on the familiar quad-rotor helicopter design, with an extra feature – the ability to stick to ceilings. That gives Eyebot the ability to linger indefinitely in advantageous positions while providing surveillance with a pan and tilt camera system. It can autonomously attach and detach itself from ceilings, and uses simple optical flow analysis to detect drift and infrared sensors to avoid obstacles.

Click here to view the embedded video.

Hand-bots are specialized in moving and acting in a space zone between the one covered by the foot-bots (the ground) and the one covered by the eye-bots (the ceiling). Hand-bots can climb vertical surfaces of walls or objects located in the environment. It’s not quite strong enough to climb using just its grippers and arm, so it has a magnetic rope launcher that can fire up to the ceiling in order to carry most of its weight.

Click here to view the embedded video.

Foot-bots are specialized in moving on rough terrain and transporting either objects or other robots. They are based on the robotic platform developed within the European Swarm-bots project. Footbot can be used in order to carry other bots and manipulate the objects on the ground. It is based on the marXbot design, which brought interesting elements as: hot swappable battery, mechanical and electronic modularity, high computational power including float processing, improved sensing ability for environment and inter-robot detection and simpler construction and maintenance.

The biggest advantage of this system is that you can easily replace or upgrade any module (robot) in the swarm. And, unlike non-modular robots, if any component fails, the swarm’s overall performance is mostly unaffected. The research will also contribute to the development of distributed algorithms meant for control of the swarmanoid and the study and definition of distributed communication protocols, thus making this and similar future projects more nimble and practical.

“You could build a lot of different robots where each does different things, but that would be too expensive,” said Wei-Min Shen, director of the Polymorphic Robotics Laboratory at the University of Southern California.

SuperBot consist of Lego-like autonomous robotic modules that can reconfigure into different systems for different tasks. Examples of configurable systems include rolling tracks or wheels (for efficient travel), spiders or centipedes (for climbing), snakes (for burrowing in ground), long arms (for inspection and repair in space), and devices that can fly in micro-gravity environment.

Click here to view the embedded video.

“Each module is a complete robotic system and has a power supply, micro- controllers, sensors, communication, three degrees of freedom, and six connecting faces (front, back, left, right, up and down) to dynamically connect to other modules.”, said Shen.

SuperBot can also demonstrate self-repair and self-assembly, because each module represents an independent robot that communicates to other modules using infrared and radio communication. The modules constantly assess the situation see what other modules are nearby, and whether they should act as an arm or a leg for the combined SuperBot.

“One of our demonstrations is that you can have a robot and cut it in half, so it becomes two independent snakes,” Shen said. “There’s no fixed central brain.”

However, a number of steps remain before SuperBot can act as a fully autonomous being that makes decisions on its own. Creating artificial intelligence (AI) that can make higher-level decisions in traditional robots has already proved challenging, and modular robots have the added complexity of deciding what represents the best shape or size in any given environment.

Click here to view the embedded video.

One solution may involve looking to biology. Shen’s lab has put forth the idea of “digital hormones” that would simulate the way in which hormones affect the mind and body. Some modules could flood certain signals to all the others, in order to communicate instructions such as whether to transform into a walking or rolling SuperBot.

The impact of SuperBot is to offer an effective and affordable solution based on modularity and re-configurability. The concepts can also be applied to other desirable system features such as reusable designs, modularity, autonomy and the easing of logistics. The interchangeability of robotic components will decrease the need for redundant parts and enhance mission reliability and safety by allowing robust re-configurability in times of failure.

Their project is a further development of the “Molecube Systems” of Cornell University, Ithaca, USA, in the third generation. The geometrical basis of this system is a cube and the Molecubes can connect themselves to others at every side. The two halves of a Molecube module rotate about the axis defined by two diagonally opposite corners. By linking together several Molecube elements, a practically unlimited number of spatial movement variants for the entire system can be realized. The end modules can also take the form of Molecubes with grippers, cameras or drive shafts.

The new configuration of an element needed to be formed is directly transferred to all the Molecubes within the system. That ensures that the energy supply and the transmission of signals between Molecubes stays preserved.

The Molecubes can be programmed in four different ways, ranging from manual to fully automated programming. High-level programming uses an interface for numerical calculations on the basis of matrices, similar to the MATLAB program, which allows programming with direct drive commands, sensor signals, and the application of internal variables and of data flow control commands. Direct programming is done via the ARM-processor interface where experienced users program the robot directly in the programming language C++. Machine learning enables fully automatic programming by means of mechanical learning processes at the highest level for research purposes is possible. Reinforcement learning and evolutionary algorithms are the key words in this context. Graphic emulation provides a realistic graphic and physical emulation allows the robot to be tested and operated virtually. In virtual reality, the user can control geometric and physical parameters and for instance monitor a robot’s collision behavior.

Click here to view the embedded video.

The developers said that the next development phase of Molecubes will be focused to drive forward mechanical and electronic integration. They also want to reduce the size of Molecubes, thus making them more versatile.

The ability to transform and adapt or even replace a modular part by using one of the functional modules that isn’t used at that moment makes transformable robots much more useable than robots with rigid structure. It is very probable that this idea will establish its application in a wide variety of new robots suitable for exploration and multitasking.