TAPAS: Throughput-adaptive Perception for Autonomous Systems

2026-07-19

AI Summary

This work addresses the inefficiency of conventional perception systems that employ fixed frame rates and static model deployments, which often lead to either resource wastage or performance degradation in dynamic environments. The authors propose a throughput-adaptive perception strategy that dynamically adjusts the target frame rate based on real-time scene complexity assessment and jointly schedules lightweight and heavyweight models on heterogeneous edge platforms. Innovatively embedding a Gated Recurrent Unit (GRU) within a Reward-based Reasoning Model (RRM) reinforcement learning framework enables end-to-end real-time resource orchestration. Experimental results on a Jetson Orin NX demonstrate that the approach achieves 93–100% throughput satisfaction with 76% energy savings on the KITTI dataset, and maintains 97% throughput satisfaction on the unseen nuScenes dataset while reducing energy consumption by 64% compared to state-of-the-art methods.

Abstract

Autonomous systems rely on a perception module to navigate through dynamic environments. In real-world scenarios, the perception module's throughput requirements vary at runtime due to changes in scene complexity. However, existing perception strategies assume a fixed FPS and static model-to-cluster mapping, resulting in either over/under provision of throughput requirements or unnecessary energy consumption across diverse scenes. Addressing this challenge requires tightly coupled \textit{scene complexity awareness} to estimate an appropriate FPS target and \textit{dynamic model-to-cluster mapping} to deliver the required throughput at minimum energy. We propose a throughput-adaptive perception strategy for mobile/edge platforms, enabling intelligent runtime resource allocation based on varying FPS targets. We use Reinforcement Learning (RL) with RRM (Reward Reasoning Model) and a GRU (Gated Recurrent Unit) agent to orchestrate perception tasks across heterogeneous mobile/edge platforms. We evaluate TAPAS on Jetson Orin NX across KITTI and unseen nuScenes. On the \textit{KITTI} dataset's test sequences, TAPAS achieves 93-100% throughput met rate while saving energy by 76%. On the unseen \textit{nuScenes} dataset, TAPAS maintains 97% throughput met rate with 64% lower energy compared to \textit{SOTA} approaches, proving its robustness.

Problem

• throughput-adaptive perception
• scene complexity awareness
• dynamic model-to-cluster mapping
• autonomous systems
• energy efficiency

Innovation

• throughput-adaptive perception
• scene complexity awareness
• dynamic model-to-cluster mapping
• reinforcement learning
• energy-efficient edge computing