Intelligence series: Swarm intelligence for robots
Latest update: March 2019
It has been suggested that swarm robotics will be able to be used for automated security systems. Additionally, NASA and DARPA have announced swarm robotics projects in development for surface planetary exploration and Unmanned Aerial Vehicle (UAV) deployment respectively.
What is Swarm Robotics?
Swarm robotics is an approach to coordinating a large number of simple robots in a distributed and decentralised manner using local rules (e.g. collision avoidance, velocity matching, flock centring).
A swarm of flying robots can maintain the same speed, direction and distance between individual units by letting each robot react to changes in its vicinity. This allows swarm systems to co-ordinate a large system of simple agents for a task that would otherwise require more advanced robots with a centralised command system. This allows swarm systems to be lower cost, fault tolerant and scalable.
‘Swarm intelligence’ or ‘swarm algorithms’ refer to a more general set of algorithms that use swarm principles to carry out a task.
What industries use Swarm Robotics?
Swarm robots can perform tasks in which the main goal is to cover wide regions or processes.
The defence industry uses swarm robotics for tasks that would otherwise be dangerous to humans, such as military reconnaissance and detecting hazardous objects such as bombs.
In the security industry, swarm robotics are being considered for automated security systems.
Emergency services use swarm robotics for surveillance in disaster rescue missions to search and access places rescue workers cannot reach. Furthermore, surveillance of areas can investigate environmental parameters, and locate sources of hazards such as chemical or gas spills, toxic pollution, pipe leaks and radioactivity.
In the agricultural industry, swarm robotics can be used for precision farming and large-scale agricultural applications. For example, they can be used for decentralised monitoring/mapping scenarios, as well as the detection and mapping of weeds/crops in a field, with a group of small Unmanned Aerial Vehicles (UAVs). This has yet to be implemented.
In the construction industry, swarm robotics are being designed to construct structures whilst being able to respond to environmental factors, such as gusts of wind.
How will it impact the rail industry?
The rail industry could implement swarm robotics in areas where processes must cover large amounts of the network. Swarm robotics can be used for monitoring of rolling stock and infrastructure, such as detection of hazards and trespass. Swarm robotics could also to support remote condition monitoring by repairing anticipated faults. Having robots covering the network would allow for on-site, predictive and preventative maintenance. Hence facilitating minimal disruption to train services.
The rail industry could also implement swarm robotics for tasks that are accelerated by having multiple robots working in parallel. This can range from simple tasks such as painting or cleaning, to more complicated tasks such as the building of new infrastructure. Swarm robotics would speed up these tasks whilst also potentially reducing expenses. This would aid in delivering low cost rail solutions.
What is the current state of R&D?
There is a large base of research on many aspects of swarm robotics. Primarily it has been focused on validation of the method through mathematical modelling, computer simulations, and robot experimentation. There has been little actual application of swarm robotic systems outside of laboratory proof of concept.
Harvard University has been researching and developing a robotic platform where block-carrying robots can build block structures much larger than themselves and soft-material-depositing robots that can build in unstructured environments. Harvard has developed the 1,024 Kilobot robot swarm to experiment with collective behaviours in large-scale autonomous swarms. This is to investigate collective ’artificial’ intelligence (e.g. sync, collective transport, self-assembly), as well as to explore new theories that link minimal individual capabilities to achievable swarm behaviours.
NASA is currently funding research into ‘Marsbees’ – a swarm of flying robots to enhance planetary surface exploration on Mars. DARPA is funding the development of ‘Gremlin drones’ – semi-disposable UAVs that can be deployed and retrieved mid-flight.
Multiple companies are presenting swarm algorithms for use in tasks such as data analysis. However, they do not apply swarm intelligence to robotics.
What uncertainties remain?
Swarm robots may be vulnerable to cyber-attacks. Attackers can implant false agents or signals to destabilise a swarm and hence hinder processes. Users must find a way to secure swarm communications whilst keeping agent cost low.
A large number of individual robots are required to make up a swarm system. This means that manufacturing and maintenance costs need to be low per unit. Current researchers cite hardware limitations as the main barrier to operational testing.
Swarm behaviour cannot be accurately predicted and so must be developed through heuristic methods. As such, it is time consuming and complex to produce robust swarm algorithms to achieve their desired task.
What should the rail industry do?
The low technology readiness level limits immediate applications of swarm robotics to the rail industry. The rail industry could collaborate with the research community and provide them with opportunities for testing in an operational environment.
Drones are already used by infrastructure managers to monitor the railway during surveys and incident management. The rail industry could apply swarm algorithms to automate these processes, hence improving coverage and surveillance uptime.
The rail industry could also apply swarm robotics to the construction and repair of infrastructure. However, if swarm robots are given access to the track, then the rail industry must evaluate how they will interact with human workers. Robots must be properly adapted so that they do not violate lineside working standards. It is important that they do not obstruct workers and put them at risk.