David Windridge, Henrik Svensson, Serge Thill

Adaptive Behavior

We consider the benefits of dream mechanisms – that is, the ability to simulate new experiences based on past ones – in a machine learning context. Specifically, we are interested in learning for artificial agents that act in the world, and operationalize “dreaming” as a mechanism by which such an agent can use its own model of the learning environment to generate new hypotheses and training data.

We first show that it is not necessarily a given that such a data-hallucination process is useful, since it can easily lead to a training set dominated by spurious imagined data until an ill-defined convergence point is reached. We then analyse a notably successful implementation of a machine learning-based dreaming mechanism by Ha and Schmidhuber (Ha, D., & Schmidhuber, J. (2018). World models. arXiv e-prints, arXiv:1803.10122). On that basis, we then develop a general framework by which an agent can generate simulated data to learn from in a manner that is beneficial to the agent. This, we argue, then forms a general method for an operationalized dream-like mechanism.

We finish by demonstrating the general conditions under which such mechanisms can be useful in machine learning, wherein the implicit simulator inference and extrapolation involved in dreaming act without reinforcing inference error even when inference is incomplete.

Read the full article at: journals.sagepub.com

Hannes Hornischer, Joshua Cherian Varughese, Ronald Thenius, Franz Wotawa, Manfred Füllsack, Thomas Schmickl

Robotic swarms and mobile sensor networks are used for environmental monitoring in various domains and areas of operation. Especially in otherwise inaccessible environments, decentralized robotic swarms can be advantageous due to their high spatial resolution of measurements and resilience to failure of individuals in the swarm. However, such robotic swarms might need to be able to compensate misplacement during deployment or adapt to dynamical changes in the environment. Reaching a collective decision in a swarm with limited communication abilities without a central entity serving as decision-maker can be a challenging task. Here, we present the CIMAX algorithm for collective decision-making for maximizing the information gathered by the swarm as a whole. Agents negotiate based on their individual sensor readings and ultimately make a decision for collectively moving in a particular direction so that the swarm as a whole increases the amount of relevant measurements and thus accessible information. We use both simulation and real robotic experiments for presenting, testing, and validating our algorithm. CIMAX is designed to be used in underwater swarm robots for troubleshooting an oxygen depletion phenomenon known as “anoxia.”

Read the full article at: journals.sagepub.com