Notch Signaling: An Agent-Based Model (ABM) of the Notch signaling pathway found in the developmental fly model, capturing in the signaling cascade as interacting populations of agents. This work is one of the first examples of using ABMs to simulate a multi-cellular as cooperative agent organizations at multiple levels. In the model, cells are composed of agents acting in an aggregate to form appropriate neural patterns to define the developing brain. [Netlogo; Java/TCL/Python; SQL] (active)

Operational Semantics of Agent-Based Modeling: Develop a statically compiled Agent-Based Modeling language that both simplifies and extends the NetLogo language. Simplification comes from a reduced core-language, while extending the agent model to allow arbitrary nesting for more complex architectures. There are two proposed targeted languages for the compiler: javascript and LLVM. [Clang++ 11; LLVM] (active)

Scouting: An automated experimentation tool based on unsupervised exploration using the concept of surprise to switch between different partially understood data niches across the global parameter space. Surprise is the comparison of an actual result to an estimate, where estimates are generated from a data-model of previous experiments. Large differences indicate poor estimates requiring additional local search with smaller parameter variance, while small differences signal a broader search and greater parameter variance. Search mechanism tends to rugged information regions, while regions of good estimates are avoided. [TCL; C++] (dormant)

Water Transport: An Agent-Based Model of potable water collection within a community without reliable water distribution. The simulation tracked family water collection, daily usage, and household water stability. Each house-hold decision making process implemented a randomized decision tree mechanism to add variance to the community decisions. [Netlogo; Java] (inactive)

Pallet Simulator: An Agent-Based Model of inter-firm wooden shipping pallet supply chain, exploring the geographic features of market structures that can evolve within it. Model insights confirm that a highly distributed and unreliable third party recyclers are capable of channeling back unclaimed pallets from a point of destination to the original shipping manufacturer. [Netlogo; Java] (inactive)

MT Learning: This work examined the Microtubule network as an information processing network within the cell, treating the system as a learning model leveraging protein assembly dynamics. Using the hypothesis that proteins are actually powerful pattern recognition devices, a early agent-oriented system was developed that combined two separate models, one based on how microtubules are assembled, and a second that treated each microtubule as a vibrating string. This system was capable of learning and generating a useful information processing structure. This work was a novel approach to modeling biological systems and illustrated how agent-oriented systems can be utilized to study systems that are not easily examined. [C++;TCL] (inactive)

Amorphous Robot: A key MtLM feature is preservation of the natural structure’s spatial relationships, allowing a reasonable fidelity to the natural system and conferring the capacity to process information in both space and time. The resulting model is particularly effective at producing tuned oscillatory behavior, making the MtLM useful for generating effective robotic gaits. To explore this behavior a biomimetic robot, called the biot, is employed that approximates the context-sensitivity inherent to biological systems. The biot consists of 12 interconnected-segments, with each segment interconnected at several points in a non-regular fashion. Placed in each segment is a servo motor that pulls/pushes each connection to produce locomotive activity through a process of twisting the biot’s structure. The effect of each motor’s movement is dependent on the current biot configuration, thus giving rise to inherent context-sensitivity at each segment that approximates the interaction of proteins found in biological systems. Structure information is retrieved from infrared-LED/photo-diode groups placed in the device at non-regular locations. Illumination levels of the different diode-groups can be sampled to provide an estimate of the current biot organization. What results is a perception-action framework, where structure information is sampled, processed relative to the MtLM internal state, and produces locomotive action. To track movement an attached computer mouse is monitored, providing current system fitness. [TCL;C++] (inactive)