[Paper Review] Impact of Heterogeneity in Multi-Robot Systems on Collective Behaviors Studied Using a Search and Rescue Problem

Sanjay Sarma O, Ramviyas Parasuraman+1 more|arXiv (Cornell University)|Dec 8, 2020

Optimization and Search Problems19 references4 citations

TL;DR

This paper investigates how heterogeneity in multi-robot systems affects collective performance in search and rescue (SAR) missions by modeling searchers and rescuers with varying functional overlap. Using Behavior Trees to quantify heterogeneity and Monte Carlo simulations in Unity, the study finds a positive correlation between heterogeneity and rescue efficiency, though excessive overlap or high-capability rescuers can reduce performance, indicating an optimal balance is needed for maximum system efficiency.

ABSTRACT

Many species in nature demonstrate symbiotic relationships leading to emergent behaviors through cooperation, which are sometimes beyond the scope of the partnerships within the same species. These symbiotic relationships are classified as mutualism, commensalism, and parasitism based on the benefit levels involved. While these partnerships are ubiquitous in nature, it is imperative to understand the benefits of collective behaviors in designing heterogeneous multi-robot systems (HMRS). In this paper, we investigate the impact of heterogeneity on the performance of HMRS applied to a search and rescue problem. The groups consisting of searchers and rescuers, varied in the individual robot behaviors with multiple degrees of functionality overlap and group compositions, demonstrating various levels of heterogeneity. We propose a new technique to measure heterogeneity in the agents through the use of Behavior Trees and use it to obtain heterogeneity informatics from our Monte Carlo simulations. The results show a positive correlation between the group's heterogeneity measure and the rescue efficiency demonstrating benefits in most of the scenarios. However, we also see cases where heterogeneity may hamper the group's abilities pointing to the need for determining the optimal heterogeneity in group required to maximally benefit from HMRS in real-world applications.

Motivation & Objective

• To investigate the impact of heterogeneity on collective performance in heterogeneous multi-robot systems (HMRS) applied to urban search and rescue (USAR).
• To develop a novel method for measuring functional heterogeneity in multi-agent systems using Behavior Trees and a distance-based metric on state-action plans.
• To evaluate how varying degrees of functionality overlap between searchers and rescuers affect system-level efficiency and cost in SAR missions.
• To determine the optimal level of heterogeneity that maximizes rescue efficiency while minimizing cost in real-world HMRS deployments.

Proposed method

• The authors model a search and rescue scenario with two robot types: searchers (locate targets) and rescuers (retrieve targets), varying their functional overlap.
• A new heterogeneity measure is proposed based on inter-species distance computed from Behavior Trees representing each robot’s state-action plan.
• Monte Carlo simulations are conducted in the Unity game engine across multiple configurations with varying ratios of searchers and rescuers.
• System performance is evaluated using time constants and a cost-weighted efficiency metric: $Efficiency = \frac{1}{\tau(c n_r + n_s)}$, where $\tau$ is average time constant and $c$ is a cost factor.
• The cost factor $c$ is calibrated based on functional disparity, with rescuers assigned higher costs than searchers.
• Simulations are run across three strategies with differing rescuer capabilities and functional overlaps to compare performance under varying heterogeneity levels.

Experimental results

Research questions

• RQ1 How does increasing functional overlap between searchers and rescuers affect the efficiency of a heterogeneous multi-robot system in a search and rescue mission?
• RQ2 Can a quantitative measure of heterogeneity be derived from behavioral representations such as Behavior Trees to predict system-level performance?
• RQ3 What is the relationship between heterogeneity level and rescue efficiency, and does it follow a unimodal trend with an optimal point?
• RQ4 How do differences in robot capabilities—especially in rescuer performance—affect the benefits of heterogeneity in team composition?

Key findings

• A positive correlation was observed between the heterogeneity measure and rescue efficiency in most scenarios, with peak performance occurring at moderate to high heterogeneity levels.
• In Scenario 3, where total population was fixed at 50, the lowest time constants (fastest retrieval) were achieved near the center of the population ratio distribution, corresponding to higher heterogeneity.
• Strategy 3, where rescuers had lower capability and cost closer to searchers, showed higher efficiency at high heterogeneity, even with blind rescuers, demonstrating the benefit of functional diversity.
• In contrast, Strategy 1, with high-performing rescuers, showed reduced efficiency when rescuers were replaced by searchers, indicating that high capability alone does not ensure better performance if it reduces functional diversity.
• The cost-weighted efficiency metric revealed that systems with lower-cost, functionally diverse robots outperformed those with high-cost, high-capability rescuers, especially when heterogeneity was balanced.
• The results suggest that optimal HMRS design requires balancing functional overlap and capability distribution, as excessive overlap or overcapacity in one agent type can diminish system-wide benefits.