Abstract:
The presence of unmanned surface vehicles (USVs) is increasingly frequent in lakes and water reservoirs, performing tasks such as monitoring water quality or delivering g...Show MoreMetadata
Abstract:
The presence of unmanned surface vehicles (USVs) is increasingly frequent in lakes and water reservoirs, performing tasks such as monitoring water quality or delivering goods across the water. However, the emergence of such autonomous vessels raises concerns in terms of safety for people sharing the same environment and the risk of collisions with fixed structures and floating bodies, including other vessels. Therefore, the detection of obstacles and their reliable operation becomes primary in USVs. This work explores the effects caused by neutron radiation on an object detection algorithm tailored for USVs. Results report 77 silent data corruption (SDC)-induced failures, showing that radiation-induced soft errors contribute to missed and false detection of, respectively, existing and nonexistent objects. Furthermore, results suggest that object detection algorithms running with the multicore strategy ( \text {FIT}_{\text {SDC}} rate of 34.3 at sea level and 308.6 at Lake Titicaca) exhibit a 16.4% greater resilience to SDCs compared to the single-core strategy.
Published in: IEEE Transactions on Nuclear Science ( Volume: 71, Issue: 8, August 2024)