Naval Artificial Intelligence Autonomous Drones – NAIAD

The main objective for the fleet is to reach a cartography area near the shore and ensuring pose uncertainties below 10m. The fleet must be quickly deployed of a fleet of autonomous UUVs at relative long distance (20km from their release area) well-positioned thanks to an associated pre-deployment of beacons dropped from a surface or aerial autonomous vehicle on the path ahead the UUVs.To do so, the vehicles must try to “hook” a beacon signal according to their local information to perform pose resetting. The hooking strategy and manoeuvre can be decided collectively by the team, a sub-team or a UUV alone. Once a beacon has been reached, the UUVs progress to the next beacon according to the plan set beforehand. This real-experiment in a marine area will be used to validate the chosen approaches. Results of the project could be useful for any scenarios where communications are disturbed (lack of GNSS in aero-terrestrial operations for example), unwanted (requited discretion), constrained by large delays or difficult in real time with operators (spatial applications). Besides the innovative scientific aspects of the proposal, the experiment involving navigation of heterogeneous multi-robot teams in a large-scale marine environment is ambitious and barely addressed in the literature.

The NAIAD project is placed in the context of mastering the underwater domain and addresses a key issue for the telecommunications (underwater optical cable) and energy (renewable energy, oil and gas and even mining resources) sectors. The NAIAD project aims to find solutions to the technological challenges related to autonomous robotic operations in an underwater environment where underwater communication has inherent limitations due to the nature of the underwater environment. Precise localisation of drones may be impossible due to the need for stealth or the absence of signals or landmarks. However, a fleet of collaborative robotic agents needs to locate itself in relation to its objectives and to exchange information in order to carry out its mission.

We therefore propose to solve the problem of deploying a fleet of underwater drones (without access to a positioning system) in a littoral zone, without immediate access for a support vessel, which remains in the open sea. The main problem in carrying out this operation is the ability for each drone to reach the area of interest and to be able to position itself accurately. To meet the need for accurate positioning, it is envisaged that an external positionable device will be used - in this case positioning beacons or daymarks dropped from an aerial or marine drone deployed from a carrier vessel sailing offshore (20km). The purpose of these synthetic landmarks will be to mark out a navigation plan in an optimal manner, taking into account the constraints of the environment and the means deployed. The dropping of these daymarks generates localization inaccuracies whose uncertainties are not perfectly known.

The aim here is to overcome the limitations caused by uncertain localisation by using automatic planning techniques in artificial intelligence that will produce robust plans, allowing to rally the daymarks even under uncertainty, while maintaining the ability to guarantee a certain degree of coordination within the fleet. Traditionally, the generation of optimised trajectories with respect to the navigation and localisation capacity of UAVs has been placed in second place with respect to technological solutions capable of ensuring better localisation; NAIAD, on the other hand, relies on trajectories aligned with strategically placed daymarks in order to favour upstream better localisation of the fleet, while aiming at the optimisation of resources (autonomy, time). The autonomy of the navigation is ensured by the execution of the plans generated by the artificial intelligence software allowing to rally the daymarks along the route. Firstly, a planner in the uncertainty will produce plans for the release of the daymarks allowing a localisation based on the daymarks and possible geographical landmarks available. The automatic synthesis of mission plans taking into account the uncertainty on the position of the UUVs will be declined according to the Temporal Hierarchical Task Network (HTN) planning paradigm, in order to split the mission into elementary tasks, which allows a rational use of resources, and to coordinate the fleet on meeting points or on the chaining of collaborative tasks. Finally, a performance monitor that evaluates the difference between the estimated position and the measured position (despite the imprecision due to signals) will act as a supervisor capable of triggering replanning episodes, in order to adapt autonomous navigation to the hazards of the environment and to be able to re-align along a navigation plan in the event of the loss of a bittern. The implementation of these planning methodologies and their integration on robotic platforms will be carried out during sea trials from the second year of the project.