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 Plus annual

Presents the access of premium content as premium feature

Premium Content

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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Deep reinforcement learning (RL) has shown immense potential for learning tocontrol systems through data alone. However, one challenge deep RL faces isthat the full state of the system is often not observable. When this is thecase, the policy needs to leverage the history of observations to infer thecurrent state. At the same time, differences between the training and testingenvironments makes it critical for the policy not to overfit to the sequence ofobservations it sees at training time. As such, there is an important balancingact between having the history encoder be flexible enough to extract relevantinformation, yet be robust to changes in the environment. To strike thisbalance, we look to the PID controller for inspiration. We assert the PIDcontroller's success shows that only summing and differencing are needed toaccumulate information over time for many control tasks. Following thisprinciple, we propose two architectures for encoding history: one that directlyuses PID features and another that extends these core ideas and can be used inarbitrary control tasks. When compared with prior approaches, our encodersproduce policies that are often more robust and achieve better performance on avariety of tracking tasks. Going beyond tracking tasks, our policies achieve1.7x better performance on average over previous state-of-the-art methods on asuite of locomotion control tasks.

PID-Inspired Inductive Biases for Deep Reinforcement Learning in Partially Observable Control Tasks