A rather new and interesting research field in mobile robotics investigates the cooperation of several units and, in particular, the possibility of obtaining a collaborative behavior by decen-tralized algorithms. Indeed, decentralized algorithms show definite advantages over centralized ones. They are naturally resilient to one-point failure (the loss of one unit), can easily reconfigure themselves, and don’t require a leader in the team. Moreover, they can exploit the increase of computing power that can be achieved by parallelizing the activities. At the same time, the number and length of interrobot communications can be kept smaller than those necessary for centralized activities. The field of decentralized algorithms can be roughly divided into two areas, the first and older one being swarm robotics, where the actions are performed by properly designed behaviors (e.g., [1]–[3]). Tuned by a trial-and-error approach, they can provide good performance and can be adapted to different robots and environments. The other area is characterized by an effort to decompose a well-posed algorithm in parts that can be individually run on different robots and provide small com- puting and communication capabilities (e.g., [4] and [5]). Under this regard, the latter area shows several similarities with that of sensor networks and can even be useful to consider each robot as a moving node of such a network. The analytical demon- stration of their convergence to the correct results is often obtained under milder assumptions than their centralized relatives and assumes far from reality characteristics such as Gaussian, zero mean errors, lossless communications, and strict interro- bot synchronization. In both cases, the results are strongly influenced by the available onboard sensors, the reliability and range of the communication system, and, obviously, the working envi- ronment. A serious assessment about the effectiveness of these algo- rithms requires a good deal of experimental work and, therefore, a team of robots. Thus, our research group decided to develop the following features in a prototype of a mobile robot: low priced, to be replicated to many copies, and at the same time, equipped with many sensors, able to communi- cate over different channels, endowed with suffi- cient computing power, and easy to program.
Di Rocco, M., La Gala, F., Ulivi, G. (2012). Testing Multirobot Algorithms: SAETTA: A Small and Cheap Mobile Unit. IEEE ROBOTICS AND AUTOMATION MAGAZINE, 20(2) [10.1109/MRA.2012.2185991].
Testing Multirobot Algorithms: SAETTA: A Small and Cheap Mobile Unit
ULIVI, Giovanni
2012-01-01
Abstract
A rather new and interesting research field in mobile robotics investigates the cooperation of several units and, in particular, the possibility of obtaining a collaborative behavior by decen-tralized algorithms. Indeed, decentralized algorithms show definite advantages over centralized ones. They are naturally resilient to one-point failure (the loss of one unit), can easily reconfigure themselves, and don’t require a leader in the team. Moreover, they can exploit the increase of computing power that can be achieved by parallelizing the activities. At the same time, the number and length of interrobot communications can be kept smaller than those necessary for centralized activities. The field of decentralized algorithms can be roughly divided into two areas, the first and older one being swarm robotics, where the actions are performed by properly designed behaviors (e.g., [1]–[3]). Tuned by a trial-and-error approach, they can provide good performance and can be adapted to different robots and environments. The other area is characterized by an effort to decompose a well-posed algorithm in parts that can be individually run on different robots and provide small com- puting and communication capabilities (e.g., [4] and [5]). Under this regard, the latter area shows several similarities with that of sensor networks and can even be useful to consider each robot as a moving node of such a network. The analytical demon- stration of their convergence to the correct results is often obtained under milder assumptions than their centralized relatives and assumes far from reality characteristics such as Gaussian, zero mean errors, lossless communications, and strict interro- bot synchronization. In both cases, the results are strongly influenced by the available onboard sensors, the reliability and range of the communication system, and, obviously, the working envi- ronment. A serious assessment about the effectiveness of these algo- rithms requires a good deal of experimental work and, therefore, a team of robots. Thus, our research group decided to develop the following features in a prototype of a mobile robot: low priced, to be replicated to many copies, and at the same time, equipped with many sensors, able to communi- cate over different channels, endowed with suffi- cient computing power, and easy to program.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
