Paolo Stegagno

Spectral analysis is a powerful techniqued used in remote sensing in order to identify and classify the earth surface. It is based on the principle that each type of surface reflects in a particular and specific way the electromagnetic radiation. This principle can be exploited by robots to identify objects of interests. Nevertheless, multispectral cameras are usually expencive and heavy. For this reason, we are developing a low-cost low-weight spectral camera suitable for the identification of vegetation.

identification is carried on individually, based on either color coding, shape identification or complex vision systems. Those methods usually assume a broad point of view over the objects, which are observed in their entirety. This assumption is sometimes difficult to fulfill in practice, and in particular in swarm systems, constituted by a multitude of small robots with limited sensing and computational capabilities. We are developing a method for target identification with a heterogeneous swarm of low-informative spatially-distributed sensors employing a distributed version of the naive Bayes classifier. Despite limited individual sensing capabilities, the recursive application of the Bayes law allows the identification if the robots cooperate sharing the information that they are able to gather from their limited points of view.

The teleoperation of a Micro Aerial Vehicle (MAV) can be employed in a great number of civil and industrial application comprising monitoring and inspection, 3D reconstruction and mapping. However, piloting this type of systems is difficult and requires many months of training. The role of the human operator can be greatly simplified by providing autonomy to the commanded robots. In particular, it is possible to: i) design innovative, simplified and intuitive ways for the human to provide command to the robots; ii) let the robots automatically avoid obstacles; iii) provide better and more informative feedback to the operator.

In order to efficiently cooperate, the component of a multi-robot team must know each other's relative pose. Although many algorithms based on reciprocal measurments exist in order to solve this problem, most of them assume that those measurements comes with the identity of the measured robots. The problem becomes more difficult when this information can not be retrieved by the sensors, leading to a combinatorial data association problem.

Localization and control in heterogeneous systems

The development of a heterogeneous system including aerial and ground robots encompasses the solution of several problems. As a prerequisite, the relative localization of the component of the team must be known. Secondly, the robots must coordinate their actions in order to fulfill the desired tasks. The characteristics of the single components can be exploited in order to improve the performance. If the aerial robots have a natural role as supervisors, hence should be equipped with rich sensor equipment, the ground robots can act as "operative hands" of the system.

Consider the problem of localizing and encircling a target using a multi-robot system. This kind of task is interesting in view of the large number of potential applications, among which we mention observation (retrieve and merge data about an object from different viewpoints), escorting (protect a member of the system from unfriendly agents) and entrapment (prevent the motion of an alien object). In this project we have developed a control scheme for achieving multi-robot encirclement in a distributed way, i.e., with each robot using only local information gathered by on-board relative-position sensors. In particular, these are assumed to be noisy, anisotropic, and unable to detect the identity of the measured object, while communication between the robots is provided by limited-range transceivers.