Reducing The Deployment-time Inference Control Costs Of Deep Reinforcement Learning Agents Via An Asymmetric Architecture

Abstract

Deep reinforcement learning (DRL) has been demonstrated to provide promising results in several challenging decision making and control tasks. However, the required inference costs of deep neural networks (DNNs) could prevent DRL from being applied to mobile robots which cannot afford high energy-consuming computations. To enable DRL methods to be affordable in such energy-limited platforms, we propose an asymmetric architecture that reduces the overall inference costs via switching between a computationally expensive policy and an economic one. The experimental results evaluated on a number of representative benchmark suites for robotic control tasks demonstrate that our method is able to reduce the inference costs while retaining the agent's overall performance.

Stats

• citations0
• S2 citations—
• github stars0
• HF likes0
• heat score0.00
• arxiv keychang2021reducing

Related papers

• Control-optimized Deep Reinforcement Learning For Artificially Intelligent Autonomous Systems (2025)0.00
• Dynamic Sparse Training For Deep Reinforcement Learning (2021)0.00
• Sequential Dynamic Decision Making With Deep Neural Nets On A Test-time Budget (2017)0.00
• Optimizing Attention And Cognitive Control Costs Using Temporally-layered Architectures (2023)2.26
• An Integrated Approach To Neural Architecture Search For Deep Q-networks (2025)0.00
• Trade-off On Sim2real Learning: Real-world Learning Faster Than Simulations (2020)3.58
• Scientific Discovery And The Cost Of Measurement -- Balancing Information And Cost In Reinforcement Learning (2021)0.00
• Efficient Off-policy Reinforcement Learning Via Brain-inspired Computing (2022)8.35