Building an MGLAIR (Modal Grounded Layered Architecture with Integrated Reasoning) Agent – Stuart Shapiro

0
5337

MGLAIR (Modal Grounded Layered Architecture with Integrated Reasoning) is a multi-layered cognitive architecture for embodied agents operating in real, virtual, or simulated environments containing other agents. The layers are organized along a mind-body axis. The highest layer, the Knowledge Layer (KL), implementing the mind, contains the beliefs of the agent, and is the layer in which conscious reasoning, planning, and act selection is performed. The lowest layer, the Sensori-Actuator Layer (SAL), contains the controllers of the sensors and effectors of the hardware or software robot. Between the KL and the SAL is the Perceptuo-Motor Layer (PML), which grounds the KL symbols in perceptual structures and subconscious actions, contains various registers for providing the agent’s sense of situatedness in the environment, and handles translation and communication between the KL and the SAL.

MGLAIR evolved from the SNePS knowledge representation and reasoning system. An acting system was added to account for the fundamental differences between acting and reasoning, and to allow for agents that combine the two. GLAIR developed as an investigation of the mind-body distinction and of mind-body coordination, especially the origin of some beliefs in perception and proprioception. MGLAIR is an extension of GLAIR with a model of concurrent multimodal sensing and acting that distributes the agents various afferent and efferent capabilities among different modalities. A single modality is a limited resource, but different modalities are independent of each other and can be used concurrently.
This two-session tutorial will build on the MGLAIR keynote and discuss the technical issues involved in building an MGLAIR agent. Among the topics to be presented are: specifying modalities; designing the PML|mind-independent sublayers, and body-independent sub-layers; perception|connecting PML structures to KL terms; primitive acts, and placing them in modalities; intensional semantics|the semantics of KL terms; the deictic center|”I”, you”, here”, and now”; the syntax and semantics of non-primitive acts; agents with and without models of time; belief revision, and why it matters.

Examples of existing MGLAIR agents will be discussed, and opportunities for further research and development of MGLAIR will be mentioned.

Stuart Shapiro presents his tutorial “Building an MGLAIR Agent” at the Sixth Conference on Artificial General Intelligence (AGI-13) in Beijing (http://www.agi-conference.org/2013).

 

 

(Source: AGI Society)

Leave a Reply