J.R. Fischer

Jörn Fischer, Thomas Ihme

Traditional AI algorithms, such as Genetic Programming and Reinforcement Learning, often require extensive computational resources to simulate real-world physical scenarios effectively. While advancements in multi-core processing have been made, the inherent limitations of parallelizing rigid body dynamics lead to significant communication overheads, hindering substantial performance gains for simple simulations. This paper introduces NeoPhysIx, a novel 3D physical simulator designed to overcome these challenges. By adopting innovative simulation paradigms and focusing on essential algorithmic elements, NeoPhysIx achieves unprecedented speedups exceeding 1000x compared to real-time. This acceleration is realized through strategic simplifications, including point cloud collision detection, joint angle determination, and friction force estimation. The efficacy of NeoPhysIx is demonstrated through its application in training a legged robot with 18 degrees of freedom and six sensors, controlled by an evolved genetic program. Remarkably, simulating half a year of robot lifetime within a mere 9 hours on a single core of a standard mid-range CPU highlights the significant efficiency gains offered by NeoPhysIx. This breakthrough paves the way for accelerated AI development and training in physically-grounded domains.

Jan Diekhoff, Jörn Fischer

Reinforcement learning is a popular method of finding optimal solutions to complex problems. Algorithms like Q-learning excel at learning to solve stochastic problems without a model of their environment. However, they take longer to solve deterministic problems than is necessary. Q-learning can be improved to better solve deterministic problems by introducing such a model-based approach. This paper introduces the recursive backwards Q-learning (RBQL) agent, which explores and builds a model of the environment. After reaching a terminal state, it recursively propagates its value backwards through this model. This lets each state be evaluated to its optimal value without a lengthy learning process. In the example of finding the shortest path through a maze, this agent greatly outperforms a regular Q-learning agent.

J. Fischer, S. Lackner

Recurrent Bistable Gradient Networks are attractor based neural networks characterized by bistable dynamics of each single neuron. Coupled together using linear interaction determined by the interconnection weights, these networks do not suffer from spurious states or very limited capacity anymore. Vladimir Chinarov and Michael Menzinger, who invented these networks, trained them using Hebb's learning rule. We show, that this way of computing the weights leads to unwanted behaviour and limitations of the networks capabilities. Furthermore we evince, that using the first order of Hintons Contrastive Divergence algorithm leads to a quite promising recurrent neural network. These findings are tested by learning images of the MNIST database for handwritten numbers.