History

The importance of environment representations for the operation of autonomous robots cannot be underestimated. Maps modeling geometrical aspects of environments, like grid maps and point clouds, are widely used in robotics. However, increasingly complex robot tasks and applications require increasingly richer representations such as semantic maps, which associate labels to portions of environments. Methods to build these maps are well covered by regular sessions at ICRA (e.g., SLAM and Semantic Scene Understanding sessions at ICRA 2022).

The goal of the workshop is to further advance in the direction of developing unconventional, possibly richer and informative, environment representations for autonomous robots, building on recent results and extending them beyond the boundaries of methods currently used in robotics.

Specifically, the workshop will reflect the state-of-the-art and stimulate innovation along the following dimensions:

  • The workshop will address both "problems" (what representation is needed for a given task) and "solutions" (what representations are available).
  • The workshop will look beyond classical explicit representations like SLAM maps, to include machine learning models that provide implicit representations, like those embedded in end-to-end neural networks that, given perceptual inputs, return environment properties or robot actions.
  • The workshop will also cover long-term representations that evolve over time to continuously adapt to the represented environments, for example, to model changing aspects or moving agents (both robots and humans).
  • The workshop will discuss what technological solutions are needed to develop suitable unconventional representations, which includes a pipeline of methods for perception, data processing, model building, and model updating.

The main expected outcome of the workshop is the definition of the first version of a "toolbox" of possible environment representations from which practitioners can choose the most suitable ones for their specific applications according to specific requirements and constraints. The availability of this toolbox is expected to stimulate research not only to expand the toolbox but also in terms of benchmarking and standardization.

History

This 2023 workshop is in continuation of
ECMR 2021 Workshop on Robotic Map Assessment and Standardisation - How good is your map? and IROS 2017 Workshop on Introspective Methods for Reliable Autonomy.