by Gil Segal, Amiram Moshaiov, Guy Amichai and Amir Ayali
Engineers are often in a need to model (identify) technical systems based on experimental data, which associate particular inputs to resulting outputs, and on insights about the structure of the system behavior. Bio-inspired system representation and learning methods, such as Artificial Neural-Networks (ANNs) and the Adaptive Neuro-Fuzzy Inference System (ANFIS) method, are often used to model technical systems by a supervised learning approach. ANNs are inspired by biological neural networks, whereas ANFIS is based on fuzzy set and fuzzy logic theories, which are inspired from the way humans perceive and describe the world. Both methods can be applied not just for the identification of technical systems, but also for the study of complex biological systems. In contrast to the use of ANNs, employing ANFIS allows building a system that is interpretable and easy to analyze. This advantage of ANFIS is expected to help understanding the complex biological system, which is dealt with here.
A locust swarm is an exceptional example of coordinated motion in nature. The general motivation for this study is the desire to understand how the behavior of the individual is translated into a collective swarm movement. This study deals with the first step towards answering this question. It concerns ANFIS-based identification of the behavior of an individual locust in a group of marching locusts, as observed in laboratory conditions.
Clearly, some major uncertainties are inherent to our understanding of the aforementioned system, as is the case with any other biological system. A major concern is the lack of knowledge about which are the inputs that a locust takes into account, when a particular motion action is experienced. Observing locusts in a swarm, it can easily be noticed that the motion is intermittent and an individual often stops for certain periods of different time-spans. We focus on the individual’s decision to initiate or resume walking, i.e. to join the collective movement. The main assumption is that if a reasonable ANFIS-based model is found for an individual locust, then it may hint at the inputs that are actually used by the locust, as related to the behavior of the swarm. Based largely on some knowledge of the biological sensory system of the locust, a trial-and-error approach has been used in order to present a bio-plausible set of inputs that provides substantially better identification results, as compared with several other such sets. The validity of the results is examined by the use of testing data, which differ from the training data.
Preliminary results suggest that the identified controller reached 85% success in predicting the behavior of the individual locust. Future work may include: a. learning the entire motion control of an individual locust, b. building a locust-inspired robot, c. testing its behavior in a swarm of real locusts, and d. providing a means for adaptation of the robot behavior in the swarm.