Hv
H.J. van Heukelum
info
Please Note
<p>This page displays the records of the person named above and is not linked to a unique person identifier. This record may need to be merged to a profile.</p>
1 records found
1
Integrated Global-Local Coverage Planning for Camera-Based ROV Monitoring of Artificial Reefs
Toward Cost-Effective and Repeatable Marine Biodiversity Surveys
Marine restoration becomes scientifically, economically, and legally defensible only when ecological change can be measured repeatably, attributed credibly, and audited. Oyster reef restoration at Project Asta in the Veerse Meer makes the difficulty concrete: clay structures seeded with oyster larvae are deployed to rebuild reefs lost over the past century, yet confirming whether these reefs come alive requires close-range underwater evidence. Diver census, drop-down video, and manually piloted Remotely Operated Vehicle (ROV) surveys are limited by safety, cost, and operator skill, and they do not guarantee consistent coverage, standoff distance, image overlap, or data quality around dense three-dimensional reef structures.
This thesis treats an ROV survey not as manually collected footage but as a constrained data-acquisition task, and asks how an integrated global-local planning approach can improve the cost-effectiveness and repeatability of camera-based ROV monitoring of artificial reef structures under operational design constraints. It contributes a framework that translates Project Asta’s monitoring needs into survey requirements and stakeholder-relevant Key Performance Indicators (KPIs), a global coverage-planning comparison for two survey tasks, a local execution layer based on Model Predictive Control (MPC)
and Model Predictive Contouring Control (MPCC), and a stakeholder-weighted benchmark that compares the proposed camera-, sonar-, and laser-equipped ROV survey workflow against conventional monitoring methods.
The framework was evaluated in simulation for a Videoray Defender-class ROV under a known environment, perfect localization, and noise-free assumptions that isolate the planning problem. Two complementary survey products were studied: a 2.5D seabed inspection intended to support a repeatable orthomosaic of the surrounding seabed, and a 3D reef-structure inspection intended to capture close-range photogrammetric imagery of each reef pile. The global planner specifies where observations are required; the local planner tests whether those references can be executed under vehicle, obstacle, and data-quality constraints. Candidate methods were assigned 0-5 scores for Data Utility,
Cost, Logistic Complexity, Mission Success Rate, and TRL, then combined with stakeholder weights for corporate actors, public authorities, NGOs, and local communities.
The seabed experiments show that the survey reference, not the choice between MPC and MPCC, governs the coverage-efficiency trade-off: the global planners reached 95.2% to 100% planned coverage, MPC and MPCC differed by less than one percentage point in executed coverage, and routing over the reef recovered boundary-adjacent patches, raising executed coverage from 95.9% to 99.9%.
The 3D reef experiments show that the global viewpoint path determines both attainable coverage and disturbance robustness: an FC-Planner-style trajectory gave the highest coverage at the shortest executed length, and MPC completed all tested constant lateral world-frame disturbance levels up to 50 N
for both evaluated references, whereas MPCC was reference-dependent, completing the FC-Plannerstyle reference up to 20 N but failing at 50 N, and failing on the 70%-overlap geodesic reference from 10 N onward.
The cost-effectiveness answer is a threshold, not an unconditional win: the proposed workflow is already the better trade-off for corporate actors and local communities, who reward higher Data Utility and reduced field effort, and it overtakes the manually piloted ROV baseline for public authorities and NGOs
only once its TRL and Mission Success Rate are raised through field validation. This crossover marks a validation gap, not a design gap. The value of the framework is that it turns operator-dependent visual acquisition into an explicit engineering object: a survey reference whose coverage, standoff, overlap, visibility, execution cost, and maturity boundary can be checked before deployment. The conclusions remain bounded by the simulation assumptions: localization error, tether loading, environmental variability, clearance, reconstruction quality, and mission success in field deployment. The most important next step is a Veerse Meer field campaign that tests whether the simulated advantages remain during field trials in the real operational design domain. ...
This thesis treats an ROV survey not as manually collected footage but as a constrained data-acquisition task, and asks how an integrated global-local planning approach can improve the cost-effectiveness and repeatability of camera-based ROV monitoring of artificial reef structures under operational design constraints. It contributes a framework that translates Project Asta’s monitoring needs into survey requirements and stakeholder-relevant Key Performance Indicators (KPIs), a global coverage-planning comparison for two survey tasks, a local execution layer based on Model Predictive Control (MPC)
and Model Predictive Contouring Control (MPCC), and a stakeholder-weighted benchmark that compares the proposed camera-, sonar-, and laser-equipped ROV survey workflow against conventional monitoring methods.
The framework was evaluated in simulation for a Videoray Defender-class ROV under a known environment, perfect localization, and noise-free assumptions that isolate the planning problem. Two complementary survey products were studied: a 2.5D seabed inspection intended to support a repeatable orthomosaic of the surrounding seabed, and a 3D reef-structure inspection intended to capture close-range photogrammetric imagery of each reef pile. The global planner specifies where observations are required; the local planner tests whether those references can be executed under vehicle, obstacle, and data-quality constraints. Candidate methods were assigned 0-5 scores for Data Utility,
Cost, Logistic Complexity, Mission Success Rate, and TRL, then combined with stakeholder weights for corporate actors, public authorities, NGOs, and local communities.
The seabed experiments show that the survey reference, not the choice between MPC and MPCC, governs the coverage-efficiency trade-off: the global planners reached 95.2% to 100% planned coverage, MPC and MPCC differed by less than one percentage point in executed coverage, and routing over the reef recovered boundary-adjacent patches, raising executed coverage from 95.9% to 99.9%.
The 3D reef experiments show that the global viewpoint path determines both attainable coverage and disturbance robustness: an FC-Planner-style trajectory gave the highest coverage at the shortest executed length, and MPC completed all tested constant lateral world-frame disturbance levels up to 50 N
for both evaluated references, whereas MPCC was reference-dependent, completing the FC-Plannerstyle reference up to 20 N but failing at 50 N, and failing on the 70%-overlap geodesic reference from 10 N onward.
The cost-effectiveness answer is a threshold, not an unconditional win: the proposed workflow is already the better trade-off for corporate actors and local communities, who reward higher Data Utility and reduced field effort, and it overtakes the manually piloted ROV baseline for public authorities and NGOs
only once its TRL and Mission Success Rate are raised through field validation. This crossover marks a validation gap, not a design gap. The value of the framework is that it turns operator-dependent visual acquisition into an explicit engineering object: a survey reference whose coverage, standoff, overlap, visibility, execution cost, and maturity boundary can be checked before deployment. The conclusions remain bounded by the simulation assumptions: localization error, tether loading, environmental variability, clearance, reconstruction quality, and mission success in field deployment. The most important next step is a Veerse Meer field campaign that tests whether the simulated advantages remain during field trials in the real operational design domain. ...
Marine restoration becomes scientifically, economically, and legally defensible only when ecological change can be measured repeatably, attributed credibly, and audited. Oyster reef restoration at Project Asta in the Veerse Meer makes the difficulty concrete: clay structures seeded with oyster larvae are deployed to rebuild reefs lost over the past century, yet confirming whether these reefs come alive requires close-range underwater evidence. Diver census, drop-down video, and manually piloted Remotely Operated Vehicle (ROV) surveys are limited by safety, cost, and operator skill, and they do not guarantee consistent coverage, standoff distance, image overlap, or data quality around dense three-dimensional reef structures.
This thesis treats an ROV survey not as manually collected footage but as a constrained data-acquisition task, and asks how an integrated global-local planning approach can improve the cost-effectiveness and repeatability of camera-based ROV monitoring of artificial reef structures under operational design constraints. It contributes a framework that translates Project Asta’s monitoring needs into survey requirements and stakeholder-relevant Key Performance Indicators (KPIs), a global coverage-planning comparison for two survey tasks, a local execution layer based on Model Predictive Control (MPC)
and Model Predictive Contouring Control (MPCC), and a stakeholder-weighted benchmark that compares the proposed camera-, sonar-, and laser-equipped ROV survey workflow against conventional monitoring methods.
The framework was evaluated in simulation for a Videoray Defender-class ROV under a known environment, perfect localization, and noise-free assumptions that isolate the planning problem. Two complementary survey products were studied: a 2.5D seabed inspection intended to support a repeatable orthomosaic of the surrounding seabed, and a 3D reef-structure inspection intended to capture close-range photogrammetric imagery of each reef pile. The global planner specifies where observations are required; the local planner tests whether those references can be executed under vehicle, obstacle, and data-quality constraints. Candidate methods were assigned 0-5 scores for Data Utility,
Cost, Logistic Complexity, Mission Success Rate, and TRL, then combined with stakeholder weights for corporate actors, public authorities, NGOs, and local communities.
The seabed experiments show that the survey reference, not the choice between MPC and MPCC, governs the coverage-efficiency trade-off: the global planners reached 95.2% to 100% planned coverage, MPC and MPCC differed by less than one percentage point in executed coverage, and routing over the reef recovered boundary-adjacent patches, raising executed coverage from 95.9% to 99.9%.
The 3D reef experiments show that the global viewpoint path determines both attainable coverage and disturbance robustness: an FC-Planner-style trajectory gave the highest coverage at the shortest executed length, and MPC completed all tested constant lateral world-frame disturbance levels up to 50 N
for both evaluated references, whereas MPCC was reference-dependent, completing the FC-Plannerstyle reference up to 20 N but failing at 50 N, and failing on the 70%-overlap geodesic reference from 10 N onward.
The cost-effectiveness answer is a threshold, not an unconditional win: the proposed workflow is already the better trade-off for corporate actors and local communities, who reward higher Data Utility and reduced field effort, and it overtakes the manually piloted ROV baseline for public authorities and NGOs
only once its TRL and Mission Success Rate are raised through field validation. This crossover marks a validation gap, not a design gap. The value of the framework is that it turns operator-dependent visual acquisition into an explicit engineering object: a survey reference whose coverage, standoff, overlap, visibility, execution cost, and maturity boundary can be checked before deployment. The conclusions remain bounded by the simulation assumptions: localization error, tether loading, environmental variability, clearance, reconstruction quality, and mission success in field deployment. The most important next step is a Veerse Meer field campaign that tests whether the simulated advantages remain during field trials in the real operational design domain.
This thesis treats an ROV survey not as manually collected footage but as a constrained data-acquisition task, and asks how an integrated global-local planning approach can improve the cost-effectiveness and repeatability of camera-based ROV monitoring of artificial reef structures under operational design constraints. It contributes a framework that translates Project Asta’s monitoring needs into survey requirements and stakeholder-relevant Key Performance Indicators (KPIs), a global coverage-planning comparison for two survey tasks, a local execution layer based on Model Predictive Control (MPC)
and Model Predictive Contouring Control (MPCC), and a stakeholder-weighted benchmark that compares the proposed camera-, sonar-, and laser-equipped ROV survey workflow against conventional monitoring methods.
The framework was evaluated in simulation for a Videoray Defender-class ROV under a known environment, perfect localization, and noise-free assumptions that isolate the planning problem. Two complementary survey products were studied: a 2.5D seabed inspection intended to support a repeatable orthomosaic of the surrounding seabed, and a 3D reef-structure inspection intended to capture close-range photogrammetric imagery of each reef pile. The global planner specifies where observations are required; the local planner tests whether those references can be executed under vehicle, obstacle, and data-quality constraints. Candidate methods were assigned 0-5 scores for Data Utility,
Cost, Logistic Complexity, Mission Success Rate, and TRL, then combined with stakeholder weights for corporate actors, public authorities, NGOs, and local communities.
The seabed experiments show that the survey reference, not the choice between MPC and MPCC, governs the coverage-efficiency trade-off: the global planners reached 95.2% to 100% planned coverage, MPC and MPCC differed by less than one percentage point in executed coverage, and routing over the reef recovered boundary-adjacent patches, raising executed coverage from 95.9% to 99.9%.
The 3D reef experiments show that the global viewpoint path determines both attainable coverage and disturbance robustness: an FC-Planner-style trajectory gave the highest coverage at the shortest executed length, and MPC completed all tested constant lateral world-frame disturbance levels up to 50 N
for both evaluated references, whereas MPCC was reference-dependent, completing the FC-Plannerstyle reference up to 20 N but failing at 50 N, and failing on the 70%-overlap geodesic reference from 10 N onward.
The cost-effectiveness answer is a threshold, not an unconditional win: the proposed workflow is already the better trade-off for corporate actors and local communities, who reward higher Data Utility and reduced field effort, and it overtakes the manually piloted ROV baseline for public authorities and NGOs
only once its TRL and Mission Success Rate are raised through field validation. This crossover marks a validation gap, not a design gap. The value of the framework is that it turns operator-dependent visual acquisition into an explicit engineering object: a survey reference whose coverage, standoff, overlap, visibility, execution cost, and maturity boundary can be checked before deployment. The conclusions remain bounded by the simulation assumptions: localization error, tether loading, environmental variability, clearance, reconstruction quality, and mission success in field deployment. The most important next step is a Veerse Meer field campaign that tests whether the simulated advantages remain during field trials in the real operational design domain.