Actryx Ltd

We provide optimization solutions for intervention problems such as those involving offline/static multi-stage and online/dynamic (real-time) decision-making.

Practical problems are often too complexity to allow for an explicit (i.e. logical) verification of the safety and optimality of the decision rules; then approximate solutions, with verification through extensive simulation and testing, has to be resorted to. By combining classical logic, machine learning, and modern simulation technology, Actryx's flexible modelling framework allows for the rigorous modelling, simulation and testing of these problems.

Our solutions take into account

• Dynamics - model dynamics causes time lags between intervention and effects. In a decision/control context this results in feedback mechanisms that must be accounted for;
• Uncertainty - any source of uncertainty in models, observations and/or measurements will require extensive simulation to generate the statistical data for quantification of optimality, safety and risk (Monte-Carlo analysis) ;
• Multi-objectivity - stakeholders have various perspectives which, at best (e.g. when stakeholders collaborate), may be slightly variant with each other; at worst (e.g. when stakeholders compete), objectives are conflicting and the problem scope takes on a game setting;
• Deadlines - hard real-time solutions require that deadlines be met. For this reason Physical Time is a first-class concept in our models so that deadlines can be tested for possible violation; if so it is an indication that the model and/or computation scheme must be modified;
• Explainability - our simulation technology produces traceable time-tagged events, underpinning the causality structure that is essential to the explainability of our solutions.

• Modularization - approximate solutions result from breaking up an infeasible problem into a hierarchy of feasible sub-problems and connecting the sub-systems (i.e. modules) in a hierarchy of coordinated interaction that guarantees the required levels of optimality, safety and performance. Modularization also allows for a systematic capability building that gradually includes more detail and complexity;
• Robust - a modularized component structure and gradual capability development process allows for a rigorous software testing policy resulting in a reliable product;
• Smooth development process - the same technolgy stack is used for rapid iterative prototyping, proof-of-concept (PoC) development and product commercialization. Thus, capitalizing on a continuity in acquiring knowledge, this allows for a smooth transition from the feasiblity (PoC) stage into the production stage.