Read Complete Research Material




Swarm robotics is an example of a complex system with interactions among distributed autonomous robots as well with the environment. Within the swarm there is no centralised control; behaviour emerges from interactions between agents within the swarm. Agents within the swarm exhibit time varying behaviour in dynamic environments, and are subject to a variety of possible anomalies. The focus within our work is on specific faults in individual robots that can affect the global performance of the robotic swarm. We argue that classical approaches for achieving tolerance through implicit redundancy are insufficient in some cases and additional measures should be explored. Our contribution is to demonstrate that tolerance through explicit detection with statistical techniques works well and is suitable due to its lightweight computation.

Table of Contents


1 Introduction3

2. Case study: a foraging swarm robotic system4

2.1. Foraging4

2.2. Forager robot4

2.3. Operational data6

2.4. Operational environment7

3. Statistical error detection8

3.1. Classical statistical techniques8

3.2. Receptor density algorithm9

4. Application of statistical classifiers for adaptive error detection11

4.1. Accessing error-detection ability12

5. Experiments and results analysis13

5.1. Experiment A14

5.2. Experiment B15

5.3. Experiment C17

6. Enhancements for improved performance19

6.1. Reducing FPR by increasing the size of the detection window19

7. Conclusions and future work21


Appendix: Referenced tables and figures26


1 Introduction

A swarm robotic system (SRS) consists of a collection of simple and homogenous miniature robots interacting with each other and the environment to perform some tasks without a centralised control [1]. It is inspired by observations in social insects which demonstrate emergent behaviour such as robustness to the lost of individuals, flexibility in carrying out tasks of different nature, and scalability in continuous operation with different group sizes [2] and [3]. These characteristics, together with distributed autonomy, make swarm robotics particularly useful for a variety of task domains such as tasks with a bounded spatial coverage, tasks that are too dangerous for human operators, tasks with dynamic scales, and tasks that require redundancy [1].

ERD involves three stages: error detection, fault diagnosis, and recovery [6] (Fig. 1). Error detection identifies erroneous states and fault diagnosis determines the causes of an error including the nature and the exact location of the faults. When a fault has been identified, recovery measures can then be carried out to prevent the faults from reoccurring. This can be done by disabling the faulty components from being invoked again. If a fault still persists after recovery measures have been carried out, that information may be used as a form of feedback to the error detection and fault diagnosis mechanism for tuning and maintenance purposes.

Fig. 1. Stages in an explicit error detection and recovery (EDR) mechanism for fault tolerance.

2. Case study: a foraging swarm robotic system

2.1. Foraging

Typically, robot foraging involves a group of robots deployed in an arena to search for specific objects and transport collected objects to a specific location [13]. At any time these robots might experience faults, (i.e. faulty motor) and the task is to detect it and recover if possible. Faults on wheels (as well as grippers) directly affect the ability of robots to forage, and we propose that the presence of these faults can be inferred, ...
Related Ads
  • College Graduates

    Swarms of start-up mature individuals, mostly ...

  • Jellyfish Swarms Debate

    In recent years, focus of the destruction caused by ...

  • Metaphor

    Accordingly, the moment when the swarm will d ...

  • Biology Questions

    Competitive displacement is accepted to be the prime ...

  • Malware

    It is possible to apply some of learning concepts de ...