Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Bayesian Optimization Framework for Efficient Fleet Design in Autonomous Multi-Robot Exploration

David Molina Concha, Jiping Li, Haoran Yin, Kyeonghyeon Park, Hyun-Rok Lee, Taesik Lee, Dhruv Sirohi, Chi-Guhn Lee

TL;DR

This work tackles the problem of designing heterogeneous autonomous robot fleets for efficient exploration under expensive, high-dimensional evaluations. It introduces BOFD, a bi-level framework that uses multi-objective Bayesian optimization to search discrete fleet compositions and a lower-level MARL method (MADDPG) to evaluate performance, while leveraging a known linear acquisition cost. A sub-linear regret bound is established to guarantee robustness, and extensive synthetic and simulated experiments show that BOFD achieves near-optimal fleet designs with far fewer evaluations than baselines. The approach promises substantial practical impact by accelerating fleet design and reducing computational costs in complex autonomous multi-robot systems.

Abstract

This study addresses the challenge of fleet design optimization in the context of heterogeneous multi-robot fleets, aiming to obtain feasible designs that balance performance and costs. In the domain of autonomous multi-robot exploration, reinforcement learning agents play a central role, offering adaptability to complex terrains and facilitating collaboration among robots. However, modifying the fleet composition results in changes in the learned behavior, and training multi-robot systems using multi-agent reinforcement learning is expensive. Therefore, an exhaustive evaluation of each potential fleet design is infeasible. To tackle these hurdles, we introduce Bayesian Optimization for Fleet Design (BOFD), a framework leveraging multi-objective Bayesian Optimization to explore fleets on the Pareto front of performance and cost while accounting for uncertainty in the design space. Moreover, we establish a sub-linear bound for cumulative regret, supporting BOFD's robustness and efficacy. Extensive benchmark experiments in synthetic and simulated environments demonstrate the superiority of our framework over state-of-the-art methods, achieving efficient fleet designs with minimal fleet evaluations.

Bayesian Optimization Framework for Efficient Fleet Design in Autonomous Multi-Robot Exploration

TL;DR

This work tackles the problem of designing heterogeneous autonomous robot fleets for efficient exploration under expensive, high-dimensional evaluations. It introduces BOFD, a bi-level framework that uses multi-objective Bayesian optimization to search discrete fleet compositions and a lower-level MARL method (MADDPG) to evaluate performance, while leveraging a known linear acquisition cost. A sub-linear regret bound is established to guarantee robustness, and extensive synthetic and simulated experiments show that BOFD achieves near-optimal fleet designs with far fewer evaluations than baselines. The approach promises substantial practical impact by accelerating fleet design and reducing computational costs in complex autonomous multi-robot systems.

Abstract

This study addresses the challenge of fleet design optimization in the context of heterogeneous multi-robot fleets, aiming to obtain feasible designs that balance performance and costs. In the domain of autonomous multi-robot exploration, reinforcement learning agents play a central role, offering adaptability to complex terrains and facilitating collaboration among robots. However, modifying the fleet composition results in changes in the learned behavior, and training multi-robot systems using multi-agent reinforcement learning is expensive. Therefore, an exhaustive evaluation of each potential fleet design is infeasible. To tackle these hurdles, we introduce Bayesian Optimization for Fleet Design (BOFD), a framework leveraging multi-objective Bayesian Optimization to explore fleets on the Pareto front of performance and cost while accounting for uncertainty in the design space. Moreover, we establish a sub-linear bound for cumulative regret, supporting BOFD's robustness and efficacy. Extensive benchmark experiments in synthetic and simulated environments demonstrate the superiority of our framework over state-of-the-art methods, achieving efficient fleet designs with minimal fleet evaluations.
Paper Structure (14 sections, 20 equations, 4 figures, 4 tables, 1 algorithm)

This paper contains 14 sections, 20 equations, 4 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Preliminaries
  4. Multi-robot exploration
  5. Partially Observable Markov Games
  6. Bayesian optimization
  7. Problem Definition
  8. Methodology
  9. Bayesian optimization for fleet design
  10. Theoretical Analysis
  11. Empirical results
  12. Synthetic Experiments
  13. Simulated Experiments
  14. Conclusions

Figures (4)

  • Figure 1: BOFD architecture using MO BO to obtain optimal fleet $\mathbf{n}$ at the upper level and the joint policy $\bf{\pi}$ is sought for the given $\mathbf{n}$ at the lower level multi-robot system.
  • Figure 2: Maps' layout for multi-robot exploration. The first row has the 2-dimensional version and the bottom row has the 3-dimensional representation.
  • Figure 3: Sensor capabilities for each of the four types of robots in a blank environment of 175x175 grids.
  • Figure 4: Average objective value $\mathcal{F}(\mathbf{n})$ and standard deviation across five priors for (a) Map #1, (b) Map #2, (c) Map #3 and (d) Map #4.