Modeling Hydra Behavior Using Methods Founded in Behavior-Based Robotics

Chapter 8Summary and conclusionsIn this thesis work methods from BBR have been used to model the behavior of a simplebiological organism. Literature studies were conducted concerning suitable model organisms,as well as methods for behavior generation and organization. Based on the literaturesurveys, the cnidarian Hydra was selected as the model organism for this project, and theCSA was selected as the framework for modeling the behavior of Hydra. The two mainreasons for choosing Hydra as the model organism are the animal’s complexity level, andthe number of measurement data available from prior ethological experiments conductedon the animal. The reasons for choosing the CSA are that it supports implementation ofpriority based behaviors (as is the case with Hydra’s behaviors according to the literature),as well as allows for modeling the reflex properties found in Hydra such as latency,habituation, and duration of response.The general conclusion of the work conducted in this project is that it is possibleto model several of the behavioral properties of Hydra using the CSA as a framework.The modeled latency of Hydra’s response to light stimulus, the habituation effect of itsresponse to mechanical stimulus, the duration of a CP and the feeding response, respectively,as well as the variable feeding threshold are consistent with experimental data.Specifically, it was found that a habituation model based on cascaded leaky integratorscan represent the habituation properties of the animal. Simulations of the overall behaviorof the animal shows that the spatial movement pattern of undisturbed Hydra, as wellas the effect of starvation on contraction frequency, agrees with results from experimentsconducted on the physical animal. In order to obtain conclusive results concerning theoverall behavior of Hydra, however, some improvements remain.Several future contributions would be of interest: Expanding the model to includethe morphology of Hydra would allow for more accurate modeling, e.g. of the durationof an LP, as well as the distance of movement, in the case of locomotion by means ofsomersaulting, as discussed in Section 7.2.1. Generation of more test data for validationpurposes would also be of interest. For example, a recording not only of the spatial45