Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Rapidly adapting robot swarms with Swarm Map-based Bayesian Optimisation

David M. Bossens, Danesh Tarapore

TL;DR

This paper tackles rapid fault recovery in swarm robotics by combining MAP-Elites-derived priors with Bayesian optimisation. It introduces SMBO, a centralised approach that uses a Gaussian process with a Matern kernel to search a behaviour-performance map, and SMBO-Decentralised, a batch asynchronous variant that assigns robots as workers with local penalisation and fault-based sharing. Empirical results on a foraging task show that SMBO and SMBO-Dec outperform random search and gradient methods, achieving substantial recovery within 30 evaluations and up to 80% improvement over pre-fault performance. The findings highlight the value of repertoire-informed priors and distributed BO for fast, robust adaptation in swarm systems, with avenues for handling dynamic environments and heterogeneous swarms.

Abstract

Rapid performance recovery from unforeseen environmental perturbations remains a grand challenge in swarm robotics. To solve this challenge, we investigate a behaviour adaptation approach, where one searches an archive of controllers for potential recovery solutions. To apply behaviour adaptation in swarm robotic systems, we propose two algorithms: (i) Swarm Map-based Optimisation (SMBO), which selects and evaluates one controller at a time, for a homogeneous swarm, in a centralised fashion; and (ii) Swarm Map-based Optimisation Decentralised (SMBO-Dec), which performs an asynchronous batch-based Bayesian optimisation to simultaneously explore different controllers for groups of robots in the swarm. We set up foraging experiments with a variety of disturbances: injected faults to proximity sensors, ground sensors, and the actuators of individual robots, with 100 unique combinations for each type. We also investigate disturbances in the operating environment of the swarm, where the swarm has to adapt to drastic changes in the number of resources available in the environment, and to one of the robots behaving disruptively towards the rest of the swarm, with 30 unique conditions for each such perturbation. The viability of SMBO and SMBO-Dec is demonstrated, comparing favourably to variants of random search and gradient descent, and various ablations, and improving performance up to 80% compared to the performance at the time of fault injection within at most 30 evaluations.

Rapidly adapting robot swarms with Swarm Map-based Bayesian Optimisation

TL;DR

This paper tackles rapid fault recovery in swarm robotics by combining MAP-Elites-derived priors with Bayesian optimisation. It introduces SMBO, a centralised approach that uses a Gaussian process with a Matern kernel to search a behaviour-performance map, and SMBO-Decentralised, a batch asynchronous variant that assigns robots as workers with local penalisation and fault-based sharing. Empirical results on a foraging task show that SMBO and SMBO-Dec outperform random search and gradient methods, achieving substantial recovery within 30 evaluations and up to 80% improvement over pre-fault performance. The findings highlight the value of repertoire-informed priors and distributed BO for fast, robust adaptation in swarm systems, with avenues for handling dynamic environments and heterogeneous swarms.

Abstract

Rapid performance recovery from unforeseen environmental perturbations remains a grand challenge in swarm robotics. To solve this challenge, we investigate a behaviour adaptation approach, where one searches an archive of controllers for potential recovery solutions. To apply behaviour adaptation in swarm robotic systems, we propose two algorithms: (i) Swarm Map-based Optimisation (SMBO), which selects and evaluates one controller at a time, for a homogeneous swarm, in a centralised fashion; and (ii) Swarm Map-based Optimisation Decentralised (SMBO-Dec), which performs an asynchronous batch-based Bayesian optimisation to simultaneously explore different controllers for groups of robots in the swarm. We set up foraging experiments with a variety of disturbances: injected faults to proximity sensors, ground sensors, and the actuators of individual robots, with 100 unique combinations for each type. We also investigate disturbances in the operating environment of the swarm, where the swarm has to adapt to drastic changes in the number of resources available in the environment, and to one of the robots behaving disruptively towards the rest of the swarm, with 30 unique conditions for each such perturbation. The viability of SMBO and SMBO-Dec is demonstrated, comparing favourably to variants of random search and gradient descent, and various ablations, and improving performance up to 80% compared to the performance at the time of fault injection within at most 30 evaluations.

Paper Structure

This paper contains 13 sections, 8 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Rapid adaptation with Swarm Map-Based Bayesian Optimisation
  3. Evolving Bayesian priors with MAP-Elites
  4. Swarm Map-based Bayesian Optimisation (SMBO)
  5. SMBO-Decentralised
  6. Experimental Setup
  7. Robot swarm and its foraging task
  8. Evolution experiments
  9. Adaptation experiments
  10. Results
  11. Evolution
  12. Adaptation
  13. Conclusion and future work

Figures (3)

  • Figure 1: Illustration of the foraging task, where 6 robots are initially positioned randomly in the arena and then must repeatedly pick up a food source and bring it to the nest. Food sources have a radius of $10cm$, $20cm$, or $30cm$ and are fixed to the illustrated positions.
  • Figure 2: Development of performance (Mean aggregated across 5 replicates and all faults within the type of perturbation) as a function of the time consumption for centralised adaptation methods.
  • Figure 3: Development of performance (Mean aggregated across 5 replicates and all faults within the type of perturbation) as a function of the time consumption for decentralised adaptation methods.