Towards Practical Infinite Stream Constraint Programming: Applications and Implementation

Siu et al. propose stream CSPs (St-CSPs) as a generalisation of finite domain CSPs to cater for constraints on infinite streams, and a solving algorithm that produces a deterministic Büchi automaton recognising the solution language. As a novel application, we demonstrate how St-CSPs can model mathematically and generate a PID controller for driving a self-balancing tray and an inverted pendulum in real-time. We propose and prove the correctness of an improvement to the implementation that eliminates numerous unnecessary states in the solution automaton for St-CSPs involving the first temporal operator, thereby reducing solving time. We give two St-CSP examples that can benefit from our new implementation techniques. Our approach always generates a solution automaton not bigger than, but potentially exponentially smaller than, that produced by the original implementation. Experimental results show substantial improvements.