Multi-Vehicle Map Fusion using GNU Radio

  • E Akin Sisbot IBM Research
  • Augusto Vega IBM Research
  • Arun Paidimarri IBM Research
  • John-David Wellman IBM Research
  • Alper Buyuktosunoglu IBM Research
  • Pradip Bose IBM Research
  • David Trilla Barcelona Supercomputing Center

Abstract

In this paper, we present a representative open-source application for fully/semiautonomous vehicles operating as a collaborative swarm using GNU Radio. This application is the driver of the DARPA-sponsored EPOCHS project led by IBM, which focuses on agile heterogeneous system-on-achip (SoC) development. The EPOCHS Reference Application (ERA) incorporates local sensing, creation of occupancy grid maps, vehicle-to-vehicle (V2V) communication of grid maps between neighboring vehicles using GNU Radio-based dedicated short-range communication (DSRC), and map fusion to create a joint higher-accuracy grid map.


This paper analyzes the GNU Radio-based DSRC transceiver workload on a Xilinx Zynq UltraScale+ MPSoC, identifying computation kernels for software and/or hardware acceleration. In particular, we present optimizations of Viterbi decoding and complex exponential that result in a measured 5:4 performance improvement over a CPU-only baseline, with projections of a fully-optimized version of this hardware accelerator showing signicantly higher benets.

Published
2019-09-17
How to Cite
SISBOT, E Akin et al. Multi-Vehicle Map Fusion using GNU Radio. Proceedings of the GNU Radio Conference, [S.l.], v. 4, n. 1, sep. 2019. Available at: <https://pubs.gnuradio.org/index.php/grcon/article/view/61>. Date accessed: 13 nov. 2019.