English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Reinforcement learning (RL) for continuous control typically employsdistributions whose support covers the entire action space. In this work, weinvestigate the colloquially known phenomenon that trained agents often preferactions at the boundaries of that space. We draw theoretical connections to theemergence of bang-bang behavior in optimal control, and provide extensiveempirical evaluation across a variety of recent RL algorithms. We replace thenormal Gaussian by a Bernoulli distribution that solely considers the extremesalong each action dimension - a bang-bang controller. Surprisingly, thisachieves state-of-the-art performance on several continuous control benchmarks- in contrast to robotic hardware, where energy and maintenance cost affectcontroller choices. Since exploration, learning,and the final solution areentangled in RL, we provide additional imitation learning experiments to reducethe impact of exploration on our analysis. Finally, we show that ourobservations generalize to environments that aim to model real-world challengesand evaluate factors to mitigate the emergence of bang-bang solutions. Ourfindings emphasize challenges for benchmarking continuous control algorithms,particularly in light of potential real-world applications.

Is Bang-Bang Control All You Need? Solving Continuous Control with Bernoulli Policies