Search for: All records

Collective behaviors in biology are thought to emerge from simple local rules, but the ecological constraints shaping these rules are difficult to test experimentally. Here, we use multi-agent reinforcement learning to train recurrent neural network agents modeled after larval zebrafish in a collective foraging task. We show that agents trained in environments with patchy food distributions spontaneously aggregate, even in the absence of food, because conspecifics serve as informative cues about likely food locations. Furthermore, penalties associated with collisions drive avoidance: stronger penalties increase preferred nearest-neighbor distances. These results provide a normative account of how learned ecological and evolutionary priors shape collective behavior in fish-like agents, establishing a virtual laboratory for testing hypotheses about the origins of collective behavior. 

Millimeter-wave (mmWave) links are increasingly utilized in wireless x-haul transport to meet growing service demands. However, the inherent susceptibility of mmWave links to weather-related attenuation creates uncertainty about future network capacity which can significantly affect Quality of Service (QoS). This creates a critical challenge: how to make admission control decisions for slices with QoS requirements, balancing acceptance rewards against the risk of future QoS-violation penalties due to capacity uncertainty? To address this, we develop a proactive slice admission control framework that tightly integrates: (i) a predictor that leverages historical link measurements to forecast short-term attenuation and quantify uncertainty; and (ii) an admission control algorithm that incorporates both the predictions and uncertainties to maximize rewards and minimize QoS-violation penalties. We compare our framework against baseline, state-of-the-art, and idealized oracle algorithms using real-world mmWave x-haul data and residential traffic traces. Simulations suggest that our framework can achieve revenues that are 250% larger than baseline algorithms and 75% larger than state-of-the-art algorithms.