Reinforcement learning competitions have formed the basis for standard research benchmarks, galvanized advances in the state-of-the-art, and shaped the direction of the field. Despite this, a majority of challenges suffer from the same fundamental problems: participant solutions to the posed challenge are usually domain-specific, biased to maximally exploit compute resources, and not guaranteed to be reproducible. In this paper, we present a new framework of competition design that promotes the development of algorithms that overcome these barriers. We propose four central mechanisms for achieving this end: submission retraining, domain randomization, desemantization through domain obfuscation, and the © W.H. Guss et al.

Holistic local visual homing based on warping of panoramic images relies on some simplifying assumptions about the images and the environment to make the problem more tractable. One of these assumptions is that images are captured on flat ground without tilt. While this might be true in some environments, it poses a problem for a wider real-world application of warping. An extension of the warping framework is proposed where tilt-corrected images are used as inputs. The method combines the tilt correction of panoramic images with a systematic search through hypothetical tilt parameters, using an image distance measure produced by warping as the optimization criterion. This method not only improves the homing performance of warping on tilted images, but also allows for a good estimation of the tilt without requiring additional sensors or external image alignment. Experiments on two newly collected tilted panoramic image databases confirm the improved homing performance and the viability of the proposed tilt-estimation scheme. Approximations of the tilt-correction image transformations and multiple direct search strategies for the tilt estimation are evaluated with respect to their runtime vs. estimation quality trade-offs to find a variant of the proposed methods which best fulfills the requirements of practical applications.