Date of Completion


Embargo Period



Underwater Swarms; Mission Planning; Target Search; Task Allocation; MiniBrain

Major Advisor

Reda A. Ammar

Associate Advisor

Sanguthevar Rajasekaran

Associate Advisor

Bing Wang

Field of Study

Computer Science and Engineering


Doctor of Philosophy

Open Access

Open Access


Modern ocean exploration and sensing approaches have been mainly based on Autonomous Underwater Vehicles (AUVs), Remotely Operated Vehicles (ROVs), and/or static Underwater Acoustic Sensor Networks (UASNs) deployments. Individual AUVs and ROVs represent a single point of failure in addition to being bulky and expensive as vehicles are usually full-featured and sophisticated. UASNs have traditionally been statically deployed. This limits their use to original deployment locations and renders them unsuitable for search tasks. Swarm Robotics (SR) are a natural, better alternative. Swarms possess superior features over a sophisticated AUV; they are smaller, cheaper, robust, reliable, and scalable by design and definition. They also have the sensing capabilities of UASNs and built-in active mobility.

Designing successful swarm missions in harsh aquatic environments is an involved task. We address this by analyzing the indispensable stages of a typical mission and carefully designing decentralized algorithms to achieve the desired per-stage goals. Important system and environmental parameters are taken into consideration to achieve the end goal: completing mission requirements while respecting time constraint, with best possible performance and minimum loss of agents. Special attention is given to target search, task identification and allocation, and mission-stage integration due to their importance. Identifying target location in an unbounded environment is challenging. Bandwidth limited and intermittent communication complicates the process further. Therefore, we develop global search algorithms that use minimal communication and utilize flocking to maintain cohesion. These algorithms have multiple advantages over traditional ones in terms of convergence time, omni-directionality, consideration of physical constraints, and being self-bounding. At the target, tasks are autonomously identified and allocated in a completely decentralized manner. Validation of the developed techniques is done through realistic simulations and analytical comparisons.

Our main contributions are: 1) a general framework for underwater mission planning, 2) three novel global search algorithms for unbounded underwater environment, 3) an algorithm for initial self-organization, 4) an optimized same-position reorientation algorithm for use in certain mission stages, 5) three autonomous task allocation algorithms, 6) three local target search algorithms, 7) a measure of mission utility, and 8) the design of a human brain-inspired model to support learning and complete autonomy.