|
|
| Available thesis |
|
Sensor fusion techniques for object recognition on a mobile robot navigating in harsh environments
The Electro Optical Sensor System (EOSS) is an electro-optical sensor system that can be mounted on a UGV to provide wide area coverage for monitoring and assessement of the area. Its main functionalities are:
• All weather area coverage
• Control of human activities at long range
• Control of vehicle movement
The EOS is installed on a power operated mount (stabilised) to permit the observation and aiming with different degree of detail also in complete darkness and in fog, misty and smoke environmental conditions. An automatic control system controls the sensors orientation, scanning a scene until locates a would-be target, which it locks onto and tracks. We propose to investigate sensor data processing algorithms in order to reliably fuse data coming from sensors, thus allowing area monitoring and assessment.
|
|
Radio communication maintenance through a multirobot system
Radio teleoperation of robotic drones normally require a robust, longrange, and non-line-of-sight (NLOS) communications link between the drone and the remote control station. High frequency digital radio communications, currently the preferred technology, are subject to line-of-sight (LOS) limitations, and thus are often impossible to maintain in complex environments (interior of a building, tunnel, or cave). We propose to solve the problem through the use of autonomous mobile relay nodes. Each node is a small robot carrying a high-bandwidth digital radio. The relay robots convoy behind the main robot at the start of a mission, and automatically stop where needed to form an ad hoc network guaranteeing a link between the lead robot and the base station. This is accomplished without the need for an operator’s intervention. Experimental tests compare the effective range achieved with and without the use of our mobile relay nodes.
|
|
3D mapping for UGV autonomous navigation in indoor environments
At Centro Gustavo Stefanini, multimodal sensor fusion algorithms are being developed in order to allow a mobile robot deal with complex environments. The navigation sensor suite is composed of a 3D Laser Range Finder and a colour CCD camera. The geometric description of the environment from the LRF has to be integrated with environment characterization from CCD camera, in order to recover the 3-D scene structure. The mapping algorithm should allow the detection of positive obstacles (that is, non-traversable areas that should be avoided), negative obstacles (such as descending stairs), and environment characteristics to help determine the most effective yet safe velocity for the traversal. While in typical 2D indoor environments the geometry description from LRF alone is sufficient to characterize the traversability of a path, 3D characteristics of the environment should also be taken into account for navigating between different floors of a building. For example, assume that the range sensors detect an obstacle 10 cm high in front of the vehicle, than this obstacle should be steered around if it is a wall or part of some other non-traversable area. However, if this obstacle is the first step of a staircase, the vehicle should recognize it and then approach to the stairs. It is apparent that only by integrating the geometry description with visual characterization will a robot be able to navigate autonomously in complex indoor environments. We propose to develop environmental features classification algorithms, based on pattern recognition techniques that should be robust to changing illumination conditions. Next, the classification is completely fused to the range data from the LRF for autonomous navigation.
|
|
|
