Airborne data collection on resilient system architectures
| Reference number | |
| Coordinator | Ericsson AB |
| Funding from Vinnova | SEK 6 054 333 |
| Project duration | May 2020 - November 2023 |
| Status | Completed |
| Venture | ECSEL |
Important results from the project
In ADACORSA, a collaboration between Ericsson, LTH, and KATAM aimed to enhance semantic SLAM. Our goal was to demonstrate the advantages of incorporating semantic information, reducing feature count, and enhancing the detectability of previously visited areas. Additionally, we explored optimizing backend calculations, minimizing memory traffic, and achieving optimal parallelization in the SLAM pipeline. The forestry scenario primarily served for the development of the semantic front end.
Expected long term effects
Our semantic front end algorithm with extended information about each landmark means that fewer landmarks are required in the calculations and an increased probability that the landmark can be rediscovered. With our toolkit for SLAM backend accelerator design, we show that our design is in line with cutting-edge research from MiT in terms of energy consumption and latency, with extended functions such as loop closure and map updates which are very computationally intensive operations. With our research steps have been taken to having full SLAM functionality in energy limited devices
Approach and implementation
The ADACORSA project had a project structure divided into work packages (WP) and supply chains (SC) to ensure that the project carried out all phases of a project from requirements, research, development to validation, verification of the sub-projects. Our project included: - Literature studies - Workshops - Development of simulation and working software to facilitate problem analysis - Analysis and improvements of hardware design in an iterative process - Test and validation of algorithm and design based on different datasets and scenarios against benchmarks - Data collection