Seminar - Professor Keith Clark - 11am Thursday 30th November 0.3 MVB
30th Nov 2017 11:00am
Title: Logic Based Tools for Engineering Ambient Intelligent Robotic Environments
Such an environment is a work place or home populated by networked sensors and controllable robotic devices that can bring about changes in the environment perhaps in collaboration with humans.
The sense data is interpreted by an intermediary layer of ‘intelligent’ software in order to determine the appropriate device responses. This software may be a collection of multi-threaded communicating and ‘thinking’ processes, called agents.
Each agent has its own beliefs and control responsibilities. They cooperate to maintain and continuously update a mutually consistent view of the current environment state and to agree on explanations of events. Periodically, a change in an agent’s beliefs may cause it to:
- issue control messages to robotic devices, or
- to notify other agents or humans of the update or its consequences, or
- to request agents or humans to perform certain tasks.
This talk with introduce a suite of logic based software and frameworks for programming such software agents. These are:
QuLog, a modern flexible typed multi-threaded logic programming language, which is much more high level than Prolog. It integrates strict functional programming, and, as in Haskell, has a clear separation between the LP+FP declarative component and an imperative component of action rules. The declarative component is used for an agent’s dynamic Beliefs and static Knowledge. The agent’s activity threads are programmed using the action rules.
- TeleoR, an extension of QuLog that is a major development of Nils Nisson’s T-R rule based language for programming robot agents. It has its roots in the triangular table representation of the action plans of the first cognitive robot, SRI’s Shakey, developed over 50 years ago. T-R and TeleoR are guarded action languages. For TeleoR the rule guards are QuLog queries to the agent’s beliefs.
- Event calculus, a logic based system for reasoning about timed events. It is used as a key component of the agent’s knowledge
- Abductive and distributed adductive reasoning. It is used to arrive at agreed explanations of events.
< Back to all events