System for cooperative perception and navigation

Purpose and goal

In this project, we will validate methods for cross-modal cooperative perception and navigation using the challenging data captured in the unstructured WARA Public Safety environments. Additionally, we will use this data to generate a benchmark involving complex localization and navigation scenarios using multiple multi-modal agents. Possible scenarios are 1) localization and navigation of ground or aerial agents exploring the outdoor and indoor environments of the research arena; 2) localization of ground agents in maps previously built by aerial vehicles and vice-versa.

Expected effects and result

The outcome of the project is two-fold. First, the creation of a public benchmarking suite covering unstructured environments, and considering seasonal changes, allowing to cover limitations of current benchmarks, e.g., KITTI, which focus only on automotive scenarios at urban areas. Second, evaluating our methods for cooperative perception and navigation as well as other relevant state-of-the-art methods. This allows us to identify the current limitations of those methods and propose future research directions and improve the design of Ericsson’s positioning solutions.

Planned approach and implementation

The project will comprise five main activities: -Identify already collected data complying with our requirements and collect new data considering seasonal changes -Process the data for the creation of the benchmark suite and define evaluation metrics -Evaluate the performance of the methods for localization and navigation -Identify limitations of the current methods in unstructured environments and their underlying causes, propose future research challenges and plans for continued collaboration -Presentation of the results and conclusions in the WASP Winter Conference