Deep Reinforcement Learning for Autonomous Underwater Navigation in Strong Ocean Currents
Train an AI agent to navigate energy-efficiently through dynamic current fields. Compare learned policies against traditional waypoint-following.
Adaptive Mission Replanning Under Degraded Sensor Conditions
Develop replanning algorithms that adjust vehicle behavior when sensors fail mid-mission. Quantify the tradeoff between mission completion and vehicle safety.
Obstacle Avoidance in Cluttered Underwater Environments Using Forward-Looking Sonar
Design and evaluate avoidance algorithms for AUVs in complex seafloor terrain using ASPEN's sonar simulation.
Autonomous Under-Ice Navigation Without GPS
Investigate navigation strategies beneath Arctic ice combining inertial navigation, acoustic positioning, and terrain-relative navigation.
Decentralized Cooperative Search Using a Swarm of Autonomous Underwater Vehicles
Design a decentralized protocol where 10–50 AUVs collaboratively search an area without centralized command.
Heterogeneous Fleet Coordination for Mine Countermeasures
Simulate a mixed fleet of long-range REMUS and short-range BlueROVs working together to detect and classify mines.
Communication-Denied Multi-Vehicle Operations
Study how a fleet maintains coordination when acoustic communication is intermittent or jammed.
Distributed Acoustic Sensing Using Coordinated Vehicle Arrays
Position multiple AUVs as a distributed acoustic sensor array to detect and localize underwater targets.
Impact of Ocean Current Forecast Uncertainty on AUV Mission Planning
Quantify how errors in current predictions affect mission outcomes using Monte Carlo simulation.
Adaptive Oceanographic Sampling
Design algorithms that direct AUVs to autonomously seek out ocean features of interest rather than following pre-programmed patterns.
Sound Velocity Profile Estimation from Mobile Platforms
Use AUV sensors to continuously estimate the SVP and evaluate impact on acoustic prediction accuracy.
Tidal Current Effects on AUV Operations in Littoral Environments
Study how tidal cycles affect mission planning using ASPEN's time-varying ocean models.
Side-Scan Sonar Coverage Optimization for Seabed Survey Missions
Optimize survey patterns to maximize sonar coverage quality while minimizing mission time.
Acoustic Communication Reliability Modeling in Shallow Water
Use Bellhop to model how environmental conditions affect acoustic comms reliability.
Terrain-Relative Navigation Accuracy as a Function of Seafloor Complexity
Evaluate TRN performance across different seafloor types using multiple real-world locations.
Optimal Vehicle-to-Mission Assignment for a Mixed AUV Fleet
Develop an optimization framework for assigning vehicles with different capabilities to concurrent missions.
Risk-Aware Path Planning Through Contested Waterways
Balance mission objectives against detection risk by modeling adversary sensor coverage as threat zones.
Battery-Constrained Mission Planning with Uncertain Energy Consumption
Build a planner that accounts for energy uncertainty due to currents and drag variations.
Rapid Environmental Assessment
Design a multi-vehicle strategy that produces an operationally useful environmental picture as quickly as possible.
Transfer Learning from Simulation to Real Underwater Vehicles
Train a policy in ASPEN, deploy on a real BlueROV2 via ROS 2, measure the sim-to-real gap.
Generative Ocean Environment Models for Simulation Diversity
Train a generative model on real oceanographic data to produce synthetic environments for training.
Digital Twin of Monterey Bay for Persistent AUV Operations
Build a high-fidelity digital twin of local waters using public data, validate against real observations.
Anomaly Detection in AUV Telemetry Using Simulation-Trained Models
Train on normal telemetry from thousands of ASPEN runs, detect anomalies during live missions.
Operator Workload in Multi-Vehicle Supervisory Control
Using the C2 UI, study how many vehicles one operator can supervise at varying autonomy levels.
Explainable Autonomy
Develop methods for AUVs to explain their decisions to operators through the C2 interface.
A Note on Scope
- All topics are scoped for 12–18 month master's theses.
- Each leverages capabilities already present in the ASPEN codebase.
- Topics can be combined or adapted to match student interests and advisor expertise.