The technical approach followed by CSI is discussed in the following along our three key re-search goals and challenges.

Multi-Perspective Situation Awareness

Assess situations across multiple perspectives.

In our project BeAware!, the view of a human operator is defined by situation type rules, basing upon various families of spatio-temporal relation types, each focusing on a particular aspect such as mereotopology. In order to apply this concept to distributed control systems, the definition of perspectives is envisioned to rest on two pillars: each perspective will provide (i) its own interpretation of relation types (e.g., “close” means “within 10km” on highways, and “at next stop” in public transportation) and (ii) a set of relevant situation types (e.g., ASFINAG operators would like to be informed whenever a “traffic jam close to road works” occurs). This means that real-world situations can be assessed by different authorities in different ways.

In order to build shared awareness about critical situations across multiple perspectives, we introduce the concept of hotspots, being spatio-temporal locations at which situations of different authorities coincide. For detecting hotspots we will develop supervised, rule-based approaches basing on relation-derivation between critical situations, and unsupervised approaches without explicit situation type rules.

Project future situations across multiple perspectives.

Qualitative projection of future (hotspot) states will be tackled by three complementary approaches: (i) Facilitating early detection of hotspots by basing upon the situation projection mechanisms proposed in BeAware! to detect whether current situations at different authorities evolve towards a hotspot, (ii) interpreting projection across authorities as a pattern recognition problem to already provide situation projections with emerging objects before they occur, and (iii) introducing evolution patterns to handle dependencies between projections of different authorities and project object evolutions irrespective of relations. Complementing our three projection approaches, projected objects of external providers, such as weather forecasts, will also be incorporated into CSI’s hotspot projection.

Scalable Collaborative Situation Awareness

Optimize awareness exchange between distributed control systems.

For optimizing awareness exchange between distributed control systems, a so-called awareness bus will be introduced, implementing a routing approach based on rules that define how authorities disseminate information on traffic objects and situations among each other. Extending such rules at runtime allows a reconfiguration of connections to share relevant information across authority boundaries.

Optimize multi-perspective situation assessment and projection.

For optimizing situation assessment and projection, reducing the number of pairwise object comparisons and possible projection paths will be tackled by both deterministic and heuristic approaches. We will employ deterministic approaches like clustering strategies examining an easily computable rule on simple object properties with respect to situation type definitions and relation type interpretations (e.g., “far” may be defined as “on different roads”; as a consequence, we need not examine objects with the same “onRoad” property value). As heuristic approaches, we plan to exploit observations such as that most objects satisfying a rule “accident causes traffic jam” are within 10km distance, instead of deterministically respecting situation type rules and relation type interpretations. We regard heuristics capable of massively reducing object comparisons, although risking that critical situations are not as-sessed (because, as in the example, there is no deterministic relationship between “causes” and “within 10km distance”). In order to reduce this risk, we will use perspective-driven se-lection of optimization techniques. For this, we first determine runtime complexity of our algorithms with respect to the number of objects in a perspective, as well as performance improvements of and dependencies between different optimization techniques. For a particular perspective, we can then estimate which techniques must be combined in order to cope with the number of objects currently active in this perspective.

Co-evolving Situation Awareness

Adapt situation assessment and projection to evolving environments.

In order to cope with ever evolving environments, we will automate the process of keeping the ontology up-to-date by (i) integrating new situation type rules, (ii) adapting relation type interpretations, and (iii) configuring projection characteristics based on real-world evolutions.

Ensure consistent situation awareness evolution across multiple perspectives.

To make the effects of changing one authority’s relation type interpretation on shared situation awareness visible, real-world information about objects as well as thereupon derived relations, assessed situations and hotspots are recorded. After changes have been made, relation derivation, situation assessment and hotspot detection is repeated on the recorded real-world objects. A consistency checker then compares recorded situations assessed on the basis of previous relation type interpretations with situations and hotspots assessed on the basis of changed interpretations. Deviations between these assessments are then presented to domain experts of different authorities, required to detect and resolve any negative impacts on situation assessment and hotspot detection in collaboration.