Indoor environments contain many linear features. We can exploit this structure to extract stable, lightweight features, to develop fast, accurate, and robust SLAM systems. That is, provided we have the right tools to deal with rank-deficient constraints.
Perceptual systems gain efficiency when they can re-use intermediate results for different inference tasks. Combining Infallible Classification and HAC, we can get multi-level inference, and perform causal analysis between hierarchy levels.
There are many SLAM systems out there, often specialized for particular environment and sensor suite configurations. Instead of creating one perfect SLAM algorithm, we are investigating if many SLAM algorithms might be used together to greater effect.
Opponent Modeling has been a field of interest for a long time. However, as more systems begin operating in challenging domains, we need a critical look at existing work and how we can push the field forward.
Explainable AI systems are in high demand as these systems proliferate beyond laboratory walls. Generating explanations is important and valuable, but naturally leads to the question: When should an agent explain?
Many environments we want robots to map are too large for a single deployment, and thus a way of combining existing maps to create a single model is necessary. This work extends our capabilities in for this task.
Copyright © 2021 Samer Nashed.