Quality-Diversity (QD) algorithms are a well-known approach to generate large collections of diverse and high-quality policies. However, QD algorithms are also known to be data-inefficient, requiring large amounts of computational resources and are slow when used in practice for robotics tasks. Policy evaluations are already commonly performed in parallel to speed up QD algorithms but have limited capabilities on a single machine as most physics simulators run on CPUs. With recent advances in simulators that run on accelerators, thousands of evaluations can performed in parallel on single GPU/TPU. In this paper, we present QDax, an implementation of MAP-Elites which leverages massive parallelism on accelerators to make QD algorithms more accessible. We first demonstrate the improvements on the number of evaluations per second that parallelism using accelerated simulators can offer. More importantly, we show that QD algorithms are ideal candidates and can scale with massive parallelism to be run at interactive timescales. The increase in parallelism does not significantly affect the performance of QD algorithms, while reducing experiment runtimes by two factors of magnitudes, turning days of computation into minutes. These results show that QD can now benefit from hardware acceleration, which contributed significantly to the bloom of deep learning.

Human‐dominated environments often include ecological traps for wildlife, such as airports that may be perceived as suitable habitat by grassland birds but reduce fitness because of collisions with aircraft. Birds of prey are often attracted to airports where collisions with aircraft (i.e., bird strikes) are usually fatal for the birds and are a significant threat to flight safety. The snowy owl (Bubo scandiacus) is known for its nomadism, exhibiting unpredictable and highly variable movements during the nonbreeding season, including being a common visitor to airports, which often have high small‐mammal populations and mimic flat, open habitats used naturally by owls. Since 2009, the Federal Aviation Administration reported an average of 22 snowy owl deaths annually due to aircraft collisions throughout 55 North American airports. To aid in active management of owls at airports, we assessed relocation data of 42 telemetry‐tracked snowy owls from 2000–2020 in the United States and Canada. Owls that returned to the airport after relocation (33%) frequently crisscrossed and perched near runways where they were at risk of strikes. Adult females and immature males were more likely to return than the other sex and age classes, and returns were less likely to occur as the distance between the release site and the airport increased. Owls relocated in open habitats with a greater proportion of wetland and cropland (including grasslands and pasture) land cover classes were also less likely to return. We conclude that inclusion of multiple factors to limit return rates of relocated snowy owls from airport facilities can unspring the ecological trap presented by airports to these owls.

No abstract

No abstract