|
|
We provide systems engineering and computational
modeling services to Professor Hunt and the other
members of the BioSystems lab. Our work for them is
supported by the CDH Research Foundation. The flagship project in this collaboration is the In Silico Liver, wherein a multi-scale collection of agents (modeling hepatic structures like sinusoids, hepatocytes, endothelial cells, enzymes, and drug molecules) act to mimic the metabolic clearance of the liver in an in situ perfusion experiment.
|
|
|
We provided custom fine-grained synthetic modeling
and simulation services to internal R&D projects to
help Bend Research better understand drug delivery in
the context of particular biological mechanisms.
|
|
|
System for Intelligent Information and Business Analysis (SIIBA) --an
"intelligent" adaptive internet-based information retrieval
and analysis application being developed for the Interactive
Institute's Tools Studio. It consists of a collection of intelligent agents, each of which is composed of a different formalism/technology: a distributed evolutionary computing-based meta-search module, a statistical module, a latent semantic analysis module, and a genetic algorithm module. Together, they act to process textual information acquired from various heterogeneous databases and distill it into actionable competitive intelligence. Here is an illustrative animation.
|
|
|
As a part of the EU Self-Organizing Innovation Networks (SEIN) project, we
designed an interesting adaptive simulation, wherein a collection of actors are specified, each with a different "kene" (akin to a genome for a genetic algorithm where the genes are various capabilities for doing some task). The actors are motivated to explore the capabilities of other actors and to cluster together to produce some artifact, which is then evaluated and determined as innovative (or not). The networks of cooperating actors arise as a consequence of the evaluation of the artifacts. The justification for the model is described in The development of a Generic Innovation Network Simulation Platform.
|
|
|
A study of the
introduction of direct-to-consumer sales of insurance via the internet
and how that new distribution channel would affect human-mediated
sales. The model consists of 3 canonical insurance companies: a traditional "brick & mortar" company that only sells through human agents, an evolutionary company that sells through both human agents and over the internet, and one that sells policies solely through the internet. The consumers are primitive Belief-Desire-Intention (BDI) actors who make their decisions based on their needs, perception of the insurance companies, and interaction with the insurance agents. The insurance agents can float between the insurance companies and, potentially, take their clients with them when they move. The insurance agents make their decisions based on their perceptions of the insurance companies and the market. The model is justified and explained in: Agent-based modeling of disrupted market ecologies: A strategic tool to think with.
|
| The DrugWar Swarm app is a toy model of three social influences on drug trafficking behavior. Each agent uses a boolean polynomial to assess (from Lefebvre's "Algebra of Consciousness") whether or not to become part of a drug cartel's supply chain operations. The inputs to the polynomial are:
- Facts -- The agent sees the world around her.
- Coercion -- The agent receives coercive pressure from law enforcement (the threat of imprisonment) and the drug cartel (the threat of bodily harm or harm to possessions or family members).
- Social Pressure -- The agent feels moral or ethical pressure from the social structures to which she belongs.
Each variable in the polynomial has a coefficient with which to modify the agent's relative weighting of the 3 inputs. The only other parametrization provided for the agents is the "awarenessNet", which is a list of aquaintances for the agent, which define the paths by which the pressures can reach her.
|
| Our ConverBots are not an example of fine-grained ABM, but instead are an example of intelligent agents. We built these artificially intelligent agents for a virtual reality exhibit built by the Interactive Institute's Tools Studio in Sweden. The whole system presents 2 VR avatars within a cave to school children who walk through the exhibit. The children could talk to the avatars via a keyboard through IRC. The avatars then responded using their AI (based on the Alice ProgramD). The responses not only show up on a monitor for the IRC channel, but are also synthesized using Festival and the audio signal is then piped to a 3D audio system. All we built were the bots, their AI, and the machinery to pipe the text through the speech synthesizer. The studio guys built the rest. (We chose not to do speech recognition because the target audience was children and inter-individual variability for recognizing speech can be tricky.)
|
|
|
