Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Distributed Multi-Robot Multi-Target Tracking Using Heterogeneous Limited-Range Sensors

Jun Chen, Mohammed Abugurain, Philip Dames, Shinkyu Park

TL;DR

This work addresses multi-robot multi-target tracking with heterogeneous, limited-range sensors by introducing a locally computable metric, normalized unused sensing capacity, to balance workload. It replaces traditional Voronoi-based partitioning with power diagrams using centroid-of-detection as the generation point and also develops a capacity-constrained Voronoi diagram (CCVD) to impose hard workload limits; both approaches are validated against baselines in ROS and MATLAB. The key finding is that heterogeneity-aware partitioning, particularly CCVD, yields significant improvements in tracking accuracy (OSPA) and workload balance as sensor diversity increases. The methods enable distributed, scalable MR-MTT with unknown, time-varying target counts, offering practical benefits for heterogeneous sensor networks in real-world scenarios.

Abstract

This paper presents a cooperative multi-robot multi-target tracking framework aimed at enhancing the efficiency of the heterogeneous sensor network and, consequently, improving overall target tracking accuracy. The concept of normalized unused sensing capacity is introduced to quantify the information a sensor is currently gathering relative to its theoretical maximum. This measurement can be computed using entirely local information and is applicable to various sensor models, distinguishing it from previous literature on the subject. It is then utilized to develop a distributed coverage control strategy for a heterogeneous sensor network, adaptively balancing the workload based on each sensor's current unused capacity. The algorithm is validated through a series of ROS and MATLAB simulations, demonstrating superior results compared to standard approaches that do not account for heterogeneity or current usage rates.

Distributed Multi-Robot Multi-Target Tracking Using Heterogeneous Limited-Range Sensors

TL;DR

This work addresses multi-robot multi-target tracking with heterogeneous, limited-range sensors by introducing a locally computable metric, normalized unused sensing capacity, to balance workload. It replaces traditional Voronoi-based partitioning with power diagrams using centroid-of-detection as the generation point and also develops a capacity-constrained Voronoi diagram (CCVD) to impose hard workload limits; both approaches are validated against baselines in ROS and MATLAB. The key finding is that heterogeneity-aware partitioning, particularly CCVD, yields significant improvements in tracking accuracy (OSPA) and workload balance as sensor diversity increases. The methods enable distributed, scalable MR-MTT with unknown, time-varying target counts, offering practical benefits for heterogeneous sensor networks in real-world scenarios.

Abstract

This paper presents a cooperative multi-robot multi-target tracking framework aimed at enhancing the efficiency of the heterogeneous sensor network and, consequently, improving overall target tracking accuracy. The concept of normalized unused sensing capacity is introduced to quantify the information a sensor is currently gathering relative to its theoretical maximum. This measurement can be computed using entirely local information and is applicable to various sensor models, distinguishing it from previous literature on the subject. It is then utilized to develop a distributed coverage control strategy for a heterogeneous sensor network, adaptively balancing the workload based on each sensor's current unused capacity. The algorithm is validated through a series of ROS and MATLAB simulations, demonstrating superior results compared to standard approaches that do not account for heterogeneity or current usage rates.
Paper Structure (30 sections, 28 equations, 12 figures, 5 tables, 2 algorithms)

This paper contains 30 sections, 28 equations, 12 figures, 5 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Distributed MR-MTT
  3. Coverage Control
  4. Sensor Heterogeneity
  5. Contribution
  6. Problem Formulation
  7. Lloyd's Algorithm
  8. PHD Filter for MTT
  9. Optimized Space Partition
  10. Normalized Unused Sensing Capacity
  11. Power Diagram Implementation
  12. Capacity-Constraint Voronoi Diagram Implementation
  13. Initial Cell Assignment
  14. Recursive Cell Assignment
  15. Optimized poses
  16. ...and 15 more sections

Figures (12)

  • Figure 1: Comparison of a Voronoi diagram (black lines), a power diagram (blue lines), and a CCVD (red curves and colored partitions). The darkness of each generation point (gray-scale dot) corresponds to its weight, i.e., power radius, with darker points having higher weights. The three diagrams converge to the same solution when the weights for all generation points are identical.
  • Figure 2: Two types of sensors used in the simulations. Type 1 and type 2 have viewing angles of $45^{\circ}$ and $240^{\circ}$, respectively. Black squares represent the location of sensors. The viewing angles, radii, and forward directions of both FoVs are indicated in the figures.
  • Figure 3: Visualization of a simulation environment in Gazebo with 5 TurtleBot3 Burger wheeled robots and multiple moving targets, represented with red balls.
  • Figure 4: Figures showing four clips during a single test using five TurtleBot3 robots, one of each type in Table \ref{['table:tb3_sensors']}, to track forty moving targets, i.e., from starting moment to [2]min [30]s (Figures \ref{['fig:0_230']}, \ref{['fig:rviz_0']}, \ref{['fig:rviz_230']}), from [3]min [40]s to [4]min [6]s (Figures \ref{['fig:340_406']}, \ref{['fig:rviz_340']}, \ref{['fig:rviz_430']}), from [4]min [30]s to [5]min [12]s (Figures \ref{['fig:430_512']}, \ref{['fig:rviz_430']}, \ref{['fig:rviz_520']}), and from [5]min [20]s to [6]min [30]s (Figures \ref{['fig:520_630']}, \ref{['fig:rviz_520']}, \ref{['fig:rviz_630']}), respectively. Figures on the first column are screenshot overlays of Gazebo GUI taken at the final time in the time interval, showing the top-view of five robots with FoVs, marked in red, and targets marked in magenta dots. Robot trajectories and target traces are also shown. Figures on the second and the third column show screenshot of RViz GUI at the beginning and the end of each clip, respectively. Red arrows show robot locations and orientations. Blue dots show target locations. Regions in different shades of grey show the assigned cells for each robot.
  • Figure 5: Figure showing trajectories over [12]min [30]s testing time of 5 robots in the square open task space.The numbers indicate the IDs of the robots.
  • ...and 7 more figures

Theorems & Definitions (7)

  • Definition 1: Normalized Unused Sensing Capacity
  • Remark 1: Choice of $\mu$
  • Remark 2
  • Definition 2: Heterogeneity Level
  • Definition 3: Total Sensing Capacity
  • Remark 3: Comparison of Results
  • Remark 4: Trade-off between algorithms