Reactive synthesis from high-level specifications that combine hard constraints expressed in Linear Temporal Logic (LTL) with soft constraints expressed by discounted-sum (DS) rewards has applications in planning and reinforcement learning. An existing approach combines techniques from LTL synthesis with optimization for the DS rewards but has failed to yield a sound algorithm. An alternative approach combining LTL synthesis with satisficing DS rewards (rewards that achieve a threshold) is sound and complete for integer discount factors, but, in practice, a fractional discount factor is desired. This work extends the existing satisficing approach, presenting the first sound algorithm for synthesis from LTL and DS rewards with fractional discount factors. The utility of our algorithm is demonstrated on robotic planning domains.
IntroductionReactive synthesis is the automated construction, from a high-level description of its desired behavior, of a reactive system that continuously interacts with an uncontrollable external environment (Church 1957).Recent applications of reactive synthesis have emerged in AI for planning and robotics tasks (Camacho, Bienvenu, and McIlraith 2019;He et al. 2019;Kress-Gazit, Lahijanian, and Raman 2018). These applications can be formulated as a deterministic turn-based interaction between a controllable system player and an uncontrollable environment player. Given a specification, the synthesis task is to generate a system strategy such that all resulting interactions with the environment satisfy the specification. A large focus in this line of work has been on synthesis from Linear Temporal Logic (LTL) specifications (Pnueli 1977;Pnueli and Rosner 1989).Yet, several desired specifications either cannot be expressed using LTL or doing is cumbersome. Examples include specifications about the quantitative properties of systems, such as rewards, costs, degrees of satisfaction, and so on. In fact, the combination of LTL with quantitative properties is used to express more nuanced and complex specifications (see Figure 1). Subsequently, synthesis algorithms from combination specifications have followed (