Export
Research Overview
Jicheng Zhao focuses on automated reasoning, logic programming, and goal specification for cognitive agents in non-deterministic environments. Early work explored translations between normal logic programs and propositional logic to improve solver efficiency. More recently, Zhao's research has addressed the development of non-monotonic temporal logics that allow for the elaboration-tolerant revision of agent goals.
Research Trajectory
Jicheng Zhao’s research trajectory, as observed in the corpus between 2003 and 2008, centers on bridging the gap between expressive high-level formalisms and efficient computational solvers, particularly in the context of planning and agent goal specification.
In 2003, Zhao co-authored On Tight Logic Programs and Yet Another Translation from Normal Logic Programs to Propositional Logic, which investigated the conditions under which the stable models of a logic program align with the propositional models of its completion. Building on prior work by Fages and others regarding program tightness, this research proposed a new normal logic program encoding for the Hamiltonian Circuit problem, demonstrating that the program remains tight on every model of its completion. This work contributed to the research direction of Modular Translation Between Logical Formalisms and the broader area of Polynomial Transformations for Efficient Logic Solving, addressing the fundamental trade-off between expressivity and solver efficiency.
By 2005, Zhao’s focus expanded to automated planning in uncertain environments. Serving as the senior author, Zhao published Using SAT and Logic Programming to Design Polynomial-Time Algorithms for Planning in Non-Deterministic Domains. This paper applied Horn SAT and logic programming techniques to develop polynomial-time algorithms for finding weak, strong, and strong cyclic plans. By utilizing non-monotonic logic programming to declaratively compute strong cyclic plans, the research provided a mechanism for singling out preferred plans among alternatives. This contribution aligned with the theme of Planning for Temporally Extended Goals, moving beyond simple point-to-point planning to handle complex, ongoing conditions in non-deterministic settings.
Following this work on planning algorithms, Zhao shifted toward the formal specification of what agents are actually expected to achieve. In 2006, Zhao co-authored Goal Specification, Non-Determinism and Quantifying over Policies. This research highlighted the limitations of existing temporal logics, such as LTL and CTL, when applied to domains where actions have non-deterministic consequences. To address this, the paper proposed P-CTL, a language incorporating quantifiers over policies, enabling more intuitive and useful goal specifications for cognitive robots.
In 2007 and 2008, Zhao’s work deepened into the area of Non-Monotonic Reasoning for Dynamic Belief Revision, specifically applying these concepts to temporal logic. The 2007 paper Non-monotonic Temporal Logics for Goal Specification introduced N-LTL, a simple extension of Linear Temporal Logic designed to handle exceptions. This allowed an agent's goals to be specified non-monotonically—for instance, maintaining a condition unless a specific exception occurs. This was followed by the 2008 paper Non-monotonic Temporal Logics that Facilitate Elaboration Tolerant Revision of Goals, which further emphasized the necessity of "elaboration tolerance." The research argued that as humans gather more knowledge about a domain, they frequently revise their requirements for an agent; therefore, formal temporal logics must support the seamless, non-monotonic revision of these goals without requiring a complete rewrite of the specification.
Current Directions
The most recent papers in the evidence, dating from 2007 and 2008, center on the formal specification and revision of goals for cognitive agents. This work investigates how non-monotonic temporal logics can support elaboration tolerance, allowing agents to adapt their objectives as they gather new information about their environments. By extending traditional frameworks like LTL to handle exceptions and dynamic belief revision, these recent contributions address the practical need for flexible, human-like goal management in non-deterministic domains.
Profile updated
Loading new papers…
Loading research neighbors…
See something wrong? Report a mistake