RL: On-Device Reinforcement Learning Runtime
Organizations: Nanyang Technological University, Singapore
Abstract
On-device reinforcement learning runs a periodic actor and a learner on one embedded GPU shared with other workloads. Because the learner produces the actor's weights, existing designs stall the actor or let its policy drift. We present RL, an on-device runtime that makes this coupling a tunable parameter: a refresh bound limits actor releases between refreshes, and a GPU permit with slack-aware admission prioritises the actor and fills spare time with training atop any base scheduler. Under the real refresh cost on two Jetson boards, RL keeps a 16.8 M-parameter policy's deadline misses below 1%, against 85 to 100% without it.
Figures & tables
| Orin Nano | Orin AGX | |||
| small | large | small | large | |
| no runtime | 71.6 / 0.6 | 11.8 / 85.3 | 90.0 / 0.4 | 30.0 / 99.9 |
| RLlib [ liang2018rllib ] | 8.1 / 3.0 | 5.5 / 1.6 | 11.9 / 1.1 | 9.8 / 3.1 |
| SF [ petrenko2020sample ] | 9.2 / 16.1 | 3.1 / 22.5 | 9.2 / 10.9 | 3.4 / 11.1 |
| RL RT | 55.1 / 0.7 | 4.7 / 0.6 | 60.0 / 0.5 | 23.7 / 0.5 |
Reinforcement Learning
Reinforcement learning studies how agents learn sequential decisions from rewards and interaction, balancing exploration, performance, safety, and adaptation.
See all 1,852 papers