Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Longitudinal Control Volumes: A Novel Centralized Estimation and Control Framework for Distributed Multi-Agent Sorting Systems

James Maier, Prasanna Sriganesh, Matthew Travers

TL;DR

This work introduces Longitudinal Control Volumes (LCV) to model material flow along a conveyor in recycling sorting facilities, enabling a low‑dimensional yet expressive state $X_k=[\mathbf{x^1_k},\dots,\mathbf{x^n_k},r_k]^T$ of dimension $nm+1$. A Kalman Filter fuses vision‑based detections into this global state, while a receding‑horizon MPC optimizes belt speed to maximize sorted value and minimize contamination over a horizon $T$, with dynamics split into motion $\mathcal{L}(r_k)$ and sorting $\mathcal{F}(\mathbf{X_k})$ operators. The approach yields substantial performance gains, with simulation and physical experiments showing $\approx 40$–$100\%$ increases in profit/value over constant‑speed baselines across 2–4 materials. The framework offers a path toward realtime centralized coordination in multi‑agent sorting and can extend to other domains, such as autonomous robots or swarm systems, where control volumes provide a scalable state representation.

Abstract

Centralized control of a multi-agent system improves upon distributed control especially when multiple agents share a common task e.g., sorting different materials in a recycling facility. Traditionally, each agent in a sorting facility is tuned individually which leads to suboptimal performance if one agent is less efficient than the others. Centralized control overcomes this bottleneck by leveraging global system state information, but it can be computationally expensive. In this work, we propose a novel framework called Longitudinal Control Volumes (LCV) to model the flow of material in a recycling facility. We then employ a Kalman Filter that incorporates local measurements of materials into a global estimation of the material flow in the system. We utilize a model predictive control algorithm that optimizes the rate of material flow using the global state estimate in real-time. We show that our proposed framework outperforms distributed control methods by 40-100% in simulation and physical experiments.

Longitudinal Control Volumes: A Novel Centralized Estimation and Control Framework for Distributed Multi-Agent Sorting Systems

TL;DR

This work introduces Longitudinal Control Volumes (LCV) to model material flow along a conveyor in recycling sorting facilities, enabling a low‑dimensional yet expressive state of dimension . A Kalman Filter fuses vision‑based detections into this global state, while a receding‑horizon MPC optimizes belt speed to maximize sorted value and minimize contamination over a horizon , with dynamics split into motion and sorting operators. The approach yields substantial performance gains, with simulation and physical experiments showing – increases in profit/value over constant‑speed baselines across 2–4 materials. The framework offers a path toward realtime centralized coordination in multi‑agent sorting and can extend to other domains, such as autonomous robots or swarm systems, where control volumes provide a scalable state representation.

Abstract

Centralized control of a multi-agent system improves upon distributed control especially when multiple agents share a common task e.g., sorting different materials in a recycling facility. Traditionally, each agent in a sorting facility is tuned individually which leads to suboptimal performance if one agent is less efficient than the others. Centralized control overcomes this bottleneck by leveraging global system state information, but it can be computationally expensive. In this work, we propose a novel framework called Longitudinal Control Volumes (LCV) to model the flow of material in a recycling facility. We then employ a Kalman Filter that incorporates local measurements of materials into a global estimation of the material flow in the system. We utilize a model predictive control algorithm that optimizes the rate of material flow using the global state estimate in real-time. We show that our proposed framework outperforms distributed control methods by 40-100% in simulation and physical experiments.
Paper Structure (13 sections, 10 equations, 9 figures)

This paper contains 13 sections, 10 equations, 9 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. Modeling Material Flow Compositions
  4. Longitudinal Control Volumes (LCV)
  5. State Transition Model
  6. Methodology
  7. State Estimation
  8. Model Predictive Control
  9. Experiments
  10. Simulation-based evaluation of MPC performance
  11. Full-system Experiments on Physical Sorting Equipment
  12. Results
  13. Conclusion and Future Work

Figures (9)

  • Figure 1: Materials on a conveyor being sorted by a human
  • Figure 2: Pictographic representation of multi-agent distributed sorting system with dynamics coupled by a single conveyor belt
  • Figure 3: Control volumes overlaid on conveyor belt serving a team of human and robot sorters
  • Figure 4: A block diagram of the proposed system to optimize belt speeds for sorting
  • Figure 5: Control volume boundaries overlaid on image with bounding box detections from object detection system. The red and blue bars at the bottom help illustrate binning for different materials and the measurement vector generated from this input image.
  • ...and 4 more figures