Liu Daxin: Probabilistic Action Formalisms with Applications to Robotics

Contact

Name

Daxin Liu

Phone

work
+49 241 80 21531

Email

E-Mail
 

The action programming language Golog was designed for the specification of the highlevel behavior of robots. Since its semantics is based on the situation calculus, a logic of action and change, it lends itself to formal investigations like the verification of temporal properties. While existing work has focused mainly on Golog programs with deterministic actions, Daxin Liu’s research considers the more realistic and challenging case of noisy actions, including noisy sensing actions. For that he has first developed an extension of a modal variant of the situation calculus with noisy actions and a probabilistic model of belief based on possible-world semantics. Using the new logic, he has considered regressionbased reasoning where queries about an agent’s beliefs after a sequence of actions can be reduced to queries about the initial beliefs of the agent. The result is very general as it can deal with arbitrary nested beliefs with quantifying-in. In ongoing work, Liu considers progression-based reasoning, where queries about the future are evaluated by first changing the agent’s knowledge base based on the effects of the actions that were performed. Liu has also begun work on the verification of Golog programs with noisy actions. Besides obaining a number of undecidability results he considers decidable fragments using approximation techniques or via a reduction to decidable cases of verifying partially observable Markov Decision Processes (POMDPs). The need for mobility and the resulting ecological, economical and social consequences are one of the key challenges for the next decades. In many rural areas the increasing urbanisation leads to a shift in mobility. The public rail transport offer is usually decreased and the private motorized transport accordingly increases its shares. This might lead to railway line discontinuations. Modern technologies allowing the use of highly automated trains and communication technology on one side and flexible transportation concepts like Dial-aRide transport on the other side can make a contribution in tackling these challenges. An on-demand rail service with small highly automated vehicles is proposed and its feasibility is researched.