# COHORT: Hybrid RL for Collaborative Large DNN Inference on Multi-Robot Systems Under Real-Time Constraints

Source: [COHORT: Hybrid RL for Collaborative Large DNN Inference on Multi-Robot Systems Under Real-Time Constraints](<https://www.sota2.com/research/paper/69b248a7d030084c22e2f9bc>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-03-13T05:00:04.336Z

Authors: Mohammad Saeid Anwar, Anuradha Ravi, Indrajeet Ghosh, Gaurav Shinde, Carl Busart, Nirmalya Roy

Published: 2026-03-11T05:38:00.000Z

## Abstract

Large deep neural networks (DNNs), especially transformer-based and multimodal architectures, are computationally demanding and challenging to deploy on resource-constrained edge platforms like field robots. These challenges intensify in mission-critical scenarios (e.g., disaster response), where robots must collaborate under tight constraints on bandwidth, latency, and battery life, often without infrastructure or server support. To address these limitations, we present COHORT, a collaborative DNN inference and task-execution framework for multi-robot systems built on the Robotic Operating System (ROS). COHORT employs a hybrid offline-online reinforcement learning (RL) strategy to dynamically schedule and distribute DNN module execution across robots. Our key contributions are threefold: (a) Offline RL policy learning combined with Advantage-Weighted Regression (AWR), trained on auction-based task allocation data from heterogeneous DNN workloads across distributed robots, (b) Online policy adaptation via Multi-Agent PPO (MAPPO), initialized from the offline policy and fine-tuned in real time, and (c) comprehensive evaluation of COHORT on vision-language model (VLM) inference tasks such as CLIP and SAM, analyzing scalability with increasing robot/workload and robustness under . We benchmark COHORT against genetic algorithms and multiple RL baselines. Experimental results demonstrate that COHORT reduces battery consumption by 15.4% and increases GPU utilization by 51.67%, while satisfying frame-rate and deadline constraints 2.55 times of the time.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Multi-robot task scheduling on Husky Robot 70-minute window](<https://www.sota2.com/research/sota/multi-robot-task-scheduling-on-husky-robot-70-minute-window.md>): Multi-robot Task Scheduling; Husky Robot 70-minute window; CPU Utilization (%): 23.8.

- [Multi-robot task scheduling on Jackal Robot 70-minute window](<https://www.sota2.com/research/sota/multi-robot-task-scheduling-on-jackal-robot-70-minute-window.md>): Multi-robot Task Scheduling; Jackal Robot 70-minute window; CPU Utilization: 25.38.

- [Multi-robot task scheduling on Spot Robot 70-minute window](<https://www.sota2.com/research/sota/multi-robot-task-scheduling-on-spot-robot-70-minute-window.md>): Multi-robot Task Scheduling; Spot Robot 70-minute window; CPU Utilization (%): 35.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2603.10436>)

[PDF](<https://arxiv.org/pdf/2603.10436.pdf>)
