Automated structural modelling for effect prediction

The outcomes can not only have a positive impact on a wide range of military activities, but may also have a dual use potential. They should in particular contribute to:

• Enhanced decision-making for operational planning activities such as targeting activities, planning of indirect fires or aerial bombings over enemy positions in urban areas, while limiting the risks of collateral damages.
• Vulnerability assessment, protection, and improvement of own infrastructure and prediction of impacts on infrastructure and operations due to e.g., natural hazards such as seismic events and tsunamis.

The effects of attacks on structures such as buildings, plants, oil tanks, pipelines, bridges, dams, etc., are a common subject of concern to military planners and engineers, weaponeers, munition designers, battle damage assessors, and modelling and simulation analysts and developers. However, predicting such effects currently involves a large uncertainty due to the difficulty in estimating the relevant characteristics of these structures. There is therefore a need to efficiently estimate such characteristics from available data such as imagery or documentation and to combine them with effect prediction models to provide reliable predictions. During operations, this should be performed in a limited time and possibly with limited available computing power. This is especially important for relatively large urban areas including many structures.

Software solutions for automated structural modelling and effect prediction should therefore be developed. They should offer the best possible accuracy, and trust should be ensured in the measurement of their performances. Given the complexity of the task, they need to rely not only on physics-based models but also on artificial intelligence, and they should be evaluated in an objective manner on data that is representative of the targeted use cases. This involves the collection and annotation of representative data. In order to ensure the reproducibility of experiments and for economic reasons, it is important that such data is reusable for similar developments, including by other technology developers. It also involves the testing of systems on new data using documented metrics and testing protocols, in a way that ensures comparability with similar systems developed by such other actors.

Models for the prediction of effects of weapons on structures are often used in conjunction with other models in decision support tools, for example to estimate freedom of manoeuvre or effectiveness of communication. Scalability and compatibility with such other models and tools should therefore be ensured.

Proposals must address the development and evaluation of software systems for modelling structures from multisource imagery and other relevant available data, and for accurately predicting the effects of weapons on these structures. This includes the collection of relevant databases for training and testing the systems.

Types of activities

The following table lists the types of activities which are eligible for this topic, and whether they are mandatory or optional (see Article 10(3) EDF Regulation):

Accordingly, the proposals must cover at least the following tasks as part of mandatory activities:

• Generating knowledge:
  • Research on automatic and semi-automatic modelling of structures from imagery and other relevant sources of information, as well as on effect prediction models, aiming at optimising the overall accuracy of effect predictions.
• Integrating knowledge:
  • Collection and annotation of representative data enabling to train and test the systems.
  • Integration of the system processing modules into demonstrators.

In addition, the proposals may also cover the following tasks:

• Generating knowledge:
  • Participation in objective and comparative evaluation campaigns or technological challenges, notably those that may be organised in the context of the EDF call topic EDF-2024-LS-RA-CHALLENGE-SPACE-MSIAO on Multi-source satellite image analysis;

The proposals should substantiate synergies and complementarities with foreseen, ongoing, or completed activities in the fields of automated structural modelling and effect prediction.

Functional requirements

The proposed product and technologies should meet the following functional requirements:

• Systems should take as input various sources of information such as satellite or aerial images (optical, IR, radar, SAR, LiDAR, etc.) and documentation if available, and estimate structural features, such as position, dimensions and composition in terms of materials of structures in specified areas. Ground and soil characteristics may also be estimated if relevant for effect prediction.
• The estimated structural models should be usable in effect prediction tools. They should enable users to estimate the degree of damage to targeted structures and their surroundings, as a function of the nature and size of the weapons used. They should also enable to estimate the potential levels of casualties depending on the nature of an attack and of the estimated human presence in a given area. Secondary effects such as window shattering should be taken into account.
• The models should lead to maximum accuracy prediction on representative databases. Reproducibility of the measurements should be ensured, by participating in existing technological challenges whenever relevant, or by organising test campaigns open to other actors if needed.
• The databases foreseen for training and testing the systems should be described in the proposals. These databases should be reusable beyond the project. The foreseen organisational and technical framework for such data sharing should be described in the proposals. In particular, the entity or entities in charge of the data production and distribution should be clearly identified in the proposals.
• Systems should allow non-expert users to evaluate effects of a certain threat over a specific target. They should also provide signatures of such targets, e.g., radar signatures. Systems should also be able to use expert user inputs in order to produce structural models in a semi-supervised manner. Demonstrators should include a user interface enabling these users to supervise the model production.
• Models and systems should be scalable and compatible with broader models supporting decision making beyond the prediction of attack effects.