Robots for Security Monitoring

: Investigating Behavioural Plasticity in Multi-Robot Systems

Abstract

This study investigates the potential benefits of incorporating behavioural plasticity into robot swarms and multi-robot systems for security monitoring applications. Drawing inspiration from biological systems, we developed and implemented a novel algorithm enabling robots to dynamically adjust their behaviour between neophilic (exploratory) and neophobic (cautious) tendencies based on environmental stimuli. The research utilised a simulated office environment designed to mirror real-world conditions, employing three Leo Rovers as robotic agents in a multi-robot system. Experiments were conducted comparing the performance of robots with plastic behaviour to those with static neophilic or neophobic temperaments. Trials ran for 1200 seconds, with a deliberate environmental change introduced at the 600-second mark to test system adaptability. Results demonstrated that robots with plastic behaviour exhibited improved performance in terms of anomaly detection, evidenced by higher aggregate ARTag detections. The plastic behaviour condition showed a demonstrable abil- ity to adapt to the simulated environmental change, maintaining higher tag collection rates compared to static behaviour conditions. Analysing dynamic "neo-threshold" values over time revealed that plastic behaviour al- lowed robots to adapt to neophilic behaviour in the initial, tag-rich environment to a neophobic tendency after the environmental change. Furthermore, the plastic behaviour condition demonstrated the lowest idleness per- centage (2.41%) compared to neophilic (6.33%) and neophobic (4.62%) conditions, indicating more efficient resource utilisation and more optimal area coverage. These findings suggest that incorporating behavioural plasticity into multi-robot systems can enhance their effectiveness in security monitoring applications, particu- larly in dynamic environments. The ability to adapt to changing conditions allowed the robots to maintain high performance levels even when anomaly frequency decreased, demonstrating a robust and flexible approach to patrolling and surveillance tasks. This research contributes to the field of swarm robotics, offering avenues for developing more adaptive and efficient autonomous security systems.

Date of Award	18 Mar 2025
Original language	English
Awarding Institution	University of Bristol
Supervisor	Paul J O'Dowd (Supervisor), R E Wilson (Supervisor) & Edmund R Hunt (Supervisor)