Objective
The Long Term objective of the project is to design the controller of an autonomous robot moving in a 3D environment and subject to external disturbances and noisy sensory data.
The specific goal of NARVAL is to increase the operational range of the platform, by designing, implementing and testing sensory-based primitives which extend the functionalities and increase the robustness of the guiding system of a mobile robot performing monitoring tasks in a partially known, unconstrained, underwater environment.
In order to operate effectively in such environmental conditions the navigational autonomy required is both at navigation level (i.e. maintain a fixed position relative to some environmental point or proceed in the direction of a sensory target) and at the observation level (i.e. completely covering a given region, or acquiring and unambiguous representation of 3D objects). At both levels visual, acoustic (sonar) and inertial sensory data will be used to control the movements of the robot and to acquire and maintain an internal representation of the environment.
Because of the intrinsic noisy nature of underwater sensory information, multi-modal sensory data will be used (e.g. vision and inertial, vision and acoustic etc.) and uncertainty will be explicitly modelled and managed to increase robustness. Relative positioning with respect to environmental data will be used for short and long-range navigation.
From the technological point of view new sensory devices will be designed and used. In particular a visio-inertial system to stabilise visual information with respect to external disturbances and to drive the low-level control of the vehicle (e.g. for station keeping). An inexpensive sonar system will also be developed providing environmental information with a lower level of detail than the vision system, but offering extended operational range and increased robustness with respect to working conditions (water turbidity, light conditions).
The work will be based on extensive use of an underwater commercial vehicle, which will be modified by the partners to allow computer control and to include new sensors. Preliminary experiments to compare different algorithmic solutions will be performed on realist 3D environments (e.g. by using Lighter Than Air blimps).
Two demonstrations will be prepared at the end of the project. The first will implement a pool-cleaning task and will demonstrate the ability of the vehicle to fully sweep the bottom of a pool using relative positioning obtained through perceptual data. The second will show a mission in a fish-farming environment and will demonstrate the ability of the vehicle to 1) return to the launching point on the basis of autonomously acquired landmarks (visual and acoustic); 2) visit in sequence a set of underwater cages and perform monitoring (e.g. observe feeding behaviour of the fishes) and maintenance action (e.g. evaluate the integrity of the cages). This last demonstration will be performed at a real fish-farming site in Théoule-sur-Mer, in the Mediterranean France coast.
Fields of science
- engineering and technologyelectrical engineering, electronic engineering, information engineeringelectronic engineeringroboticsautonomous robots
- natural sciencesphysical sciencesacoustics
- engineering and technologyelectrical engineering, electronic engineering, information engineeringelectronic engineeringsensors