Robust autonomy

Robust Autonomy (RA) is an ability of the system to react on both explicitly-specified adverse situations and unexpected adverse situations in both the environment and the underlying system (hardware and software) without or with a limited human guidance in order to complete its mission in the best possible (acceptable) way. It is the potential of an intelligent man-made mechanism for detection and reasoning about adverse situations without human intervention. The robustness of robot autonomy should be such that interactions with the environment based on reasoning remain sufficiently safe, accurate, etc. Therefore BRICS concentration in this area is on methodology of elegantly enriching the regular behavioural model with the add-ons dealing with the robust autonomy of robot. This methodology will consist of steps in robot development process and robust autonomy add-ons.

Robot development should be structured in a way that reveals decisions that affect non-functional requirements such as robust autonomy. The engineering process can be described as a combination of design procedures and artefacts (deliverables). An artefact is the most explicit way to ensure fulfilment of the requirements. Therefore, it is good practice to include robust autonomy requirements in it. BRICS will come up with an ontology for robust autonomy from which a user (designer) can derive which criteria should be taken into account in a specific context. This will help the user to make decisions about the particular demands of the context at hand and to limit the set of possible solutions.

Robust autonomy solutions could be viewed as features included in robot − resources that ensure "the happy flow" of its operation. Utilizing or introducing some type of redundancy into the system could achieve this. An example of utilizing informational redundancy is a model of the desired behaviour. In the extreme case that the system is described as a "black box", so the model merely consist of its specifications, it is only possible to detect potential failure, but not to resolve it. The more relations are known between the specifications and the internal functions, the more options there are to adapt these functions when parts are failing, such that the specifications are met according to the required level of quality (quality of service), which in turn, can be used as a criterion for robustness.

The only time when an autonomous robot can be provided with resources to achieve robust autonomy is at design time. It is essential to have structured robot development process as well as give robotic appliance all possible resources to ensure robust autonomy of the final design.

To reach the goals of this activity the following contributions are planned: