Home -> Research -> Reasoning about Action

Reasoning about action concerns how to specify and verify dynamic systems, which plays a central role in many areas of AI including planning, cognitive robotics, decision making, and so on. On the other hand, AI tends to describe and explain human behavior through mental states representing beliefs, desires, and intentions. The Belief-Desire-Intention (BDI) paradigm, largely used in AI, states that individual decisions arise from the recursive exchange of information between these three mental states.

Our group addresses both issues. We follows the style of situation calculus, focuses on specifying actions and their (direct/indirect) effects in McCann and Turner's nonmonotonic causal theories. We are interested in formalizing causal knowledge for representing action domains of service robots in domestic applications and tackling encountered problems including integration with other cognitive functions, and increasing the computing efficiency of causal reasoning. We also consider representing and reasoning about motivational attitudes about desires. We apply these research results on our service robot "KeJia" to develop plans and keep track of changes to achieve certain goals.

Recent Projects
    Integrating Reasoning about Action with other High-level Cognitive Functions for Service Robots
        Reasoning about action is a central requirement of an autonomous agent trying to achieve some goals. Meanwhile, in order to provide services for untrained and non-technical users in the domestic environment, a service robot requires having appropriate cognitive abilities, including natural language processing, dialogue understanding, task planning, learning, and so on. How to construct a cohesive architecture integrating reasoning about action with other cognitive functions becomes a challenge problem.

        To address the problem, we have proposed a layered architecture which extracts users' requirements and information from human-robot dialog and natural language processing, transfers them into the task planning component, then sends computed high-level plans to the motion planning and robot control components. Task planning is the central component of the architecture. It is asked to understand corresponding information and generate plans within a given time limit to accomplish users' requirements, which require formalizing the action domain for reasoning about action. The architecture can later be generated to integrate other cognitive functions such as learning. We have implemented the architecture and corresponding mechanisms on our service robot, "KeJia".

    Increasing the Computing Efficiency of Causal Theories for Reasoning about Action
        Causal theories are closely related to Answer Set Programming (ASP). There exists a polynomial translation from causal theories to ASP. Thus ASP solvers can be used to compute causally explained interpretations of a causal theory. On the other hand, deciding whether an interpretation is causally explained is \Sigma_2^P complete for causal theories and NP-complete for definite causal theories. How to improve the computing efficiency of causal theories is a challenge towards real applications.

        We propose an approach to speed up current ASP solvers via some consequences of a program that can be computed efficiently. A consequence of a logic program under answer set semantics is one that is true in all answer sets. We consider using loop formulas of loops with at most one external support rule through an interactive procedure using unit propagation to compute some of these consequences to increase the efficiency of answer set solvers. Corresponding algorithms have been implemented, and our experiments show that for certain logic programs, the consequences computed by our algorithms can significantly speed up current ASP solvers cmodels, clasp, and DLV.

    Reasoning about Implicit Desire
        Motivational attitudes play an important role in investigations into intelligent agents. One of the key problems of representing and reasoning about motivational attitudes is which propositions are the desirable ones. The answer based on classical logic is that the propositions that logically imply the goal are desirable and the others are not. We argue that this criterion is inadequate for the incomplete knowledge about environments for an agent. We present a simple and intuitive semantics---partial implication---for the characterization of desirable propositions. In this semantics, Proposition P is a desirable one with respect to a given goal Q if and only if P is "useful" and "harmless" to Q in any situation. Partial implication is an extension of classical implication. We investigate some fundamental properties of partial implication and discuss some of the potential applications.

    Xiaoping Chen, Jianmin Ji, Fangzhen Lin. Computing loops with at most one external support rule. ACM Transactions on Computational Logic (TOCL). 2012 (To appear). .
    Xiaoping Chen, Jianmin Ji, Jiehui Jiang, Guoqiang Jin, Feng Wang, Jiongkun Xie. Developing High-level Cognitive Functions for Service Robots, In: Proc. of 9th Int. Conf. on Autonomous Agents and Multi-agent Systems (AAMAS 2010), Toronto, Canada, May 2010. .
    Xiaoping Chen, Jiehui Jiang, Jianmin Ji, Guoqiang Jin and Feng Wang. Integrating NLP with reasoning about actions for autonomous agents. The Proceedings of the 2009 IEEE/WIC/ACM International Joint Conferences on Web Intelligence and Intelligent Agent Technologies (WI-IAT 2009) Milano, Italy, Sep. 2009. .
    Xiaoping Chen and Jianmin Ji and Fangzhen Lin. Computing Loops with at Most One External Support Rule for Disjunctive Logic Programs. Proceedings of 25th International Conference on Logic Programming (ICLP' 09), Pasadena, California, July, 2009. .
    Xiaoping Chen, Jianmin Ji, Fangzhen Lin. Computing Loops with at Most One External Support Rule. Proceedings of KR 08, Sydney, Australia, September 16-19, 2008. .
    Zhou Yi and Chen Xiaoping. Formalizing probabilistic relevance in propositional language. Journal of University of Science and Technology of China, Vol. 37, NO 12, Dec. 2007. .
    Yi Zhou and XiaoPing Chen. Toward formalizing usefulness in propositional language. In: J. Lang, F. Lin, and J. Wang (Eds.). The Proceedings of the First International Conference on Knowledge Science, Engineering and Management (KSEM'06), LNAI 4092, Springer, 2006. .
    Yi Zhou and Xiaoping Chen. Implicit desire and its formulization. Journal of Software, 16(5):771-778, 2005. .
    Yi Zhou and Xiaoping Chen. Partial implication semantics for desirable propositions. In: Proceedings of KR-04, Aug. 2004. .