EUROPtus: A Mixed-Initiative Controller for Multi-vehicle Oceanographic Field Experiments

Review of mission planning for autonomous marine vehicle fleets

“…It was concluded that the network produced effective coordination between aerial, surface and submerged resources, but required direct communication with the operator to achieve this result. In the latest trials of the LSTS toolchain, a mixed‐initiative planner called EUROPtus (Py et al, ) was implemented on a fleet of AUVs and UAVs to perform cetacean monitoring (similar to the mission objectives in Bennett et al ()). The centralised, deliberative planner can decompose abstract operator tasks and schedule them to nominated vehicles, allowing the operator to focus on risk analysis and sampling strategies for each individual vehicle.…”

“…The current standard in heterogeneous AMV fleet planning is vehicle‐by‐vehicle task planning using EUROPtus, a centralised, mixed‐initiative temporal planner (Py et al, ). Mixed‐initiative planning is powerful because it allows the operator to quickly describe custom tasks for the planner to schedule and allocate, making it adaptable to planning many kinds of science missions.…”

“…If the assigned nodes have been allocated a set of dependent, but dissimilar tasks, then they are acting with a collaborative behaviour. The heterogeneous fleets deployed in (Faria et al, ; Pinto et al, ; Py et al, ) for marine sampling cotemporally with different sensory payloads on each vehicle are excellent examples. A collaborative mission that involves a more obvious dependency is in the detection and then disarming of mines as addressed by Sotzing and Lane ().…”

“…For a centralised planning schema to work effectively, communication between the acting agents and the planning agent must be reliable and regularly available so that the risk of either of the mentioned planning failures is sufficiently low. The EUROPtus planner (Py et al, ) is a centralised planner designed to schedule and allocate abstract tasks for individual vehicles. Each vehicle’s onboard local planner further decomposes these tasks into a schedule of vehicle specific behaviours.…”

“…In fleet deployments, the planning space is scaled up in complexity as each vehicle contributes its own action‐state space to the mission. The T‐REX planner does not perform efficiently on larger scale planning problems (Ingrand & Ghallab, ) and has been supplemented by a centralised, shore‐side, mixed‐initiative planner (Py et al, ) to formulate goals for the individual T‐REX planners onboard each vehicle.…”