Managed Swarm Robotics for Autonomous Disaster Response - The Concept
- The Founder
- Jul 14
- 6 min read
In this project, I explore the concept of an autonomous disaster response system composed of low-cost, expendable robots. This formed the basis of my Computer Systems Engineering bachelor’s thesis, "Integrating Object Recognition in Swarm Robotics Management," at Brunel University in 2022.
Summary
In a world with so many prominent large-scale disasters, I envisioned a system that could act autonomously to deploy robots to intercept the hazards upon detection. The motivation was to reduce response time whilst, more importantly, limiting both the risk towards and the dependency of human actors. This concept is what I sought to explore during my Computer Systems Engineering bachelor's thesis, "Integrating Object Recognition in Swarm Robotics Management", back in 2022.
The aim was to develop a proof of concept system based on simplistic hardware to investigate and demonstrate how an autonomous ecosystem could detect, locate and mitigate hazards without a human on-site presence. This includes a fleet of semi-autonomous robots with simplistic control hardware and an overlooking system that can detect a hazard and command a robot to take action accordingly.
Introduction
To overcome a disaster scenario, a vast amount of time-critical hazards must be resolved in order to prevent the situation from worsening. This is also done whilst attempting to rescue people and protect property affected by the situation. This, of course, depends on taking action that directly puts respondents at risk and, as seen in history, includes tasks that can easily prove fatal.
During my research, I found many examples of robots being utilised to undertake crucial tasks within previous disaster scenarios. This includes the clearing of radioactive debris at Chernobyl, putting out fires at the Notre Dame Cathedral, and even the use of drones to map and project the spread of forest wildfires. The successes and failures of these historical approaches helped form my approach for a system that may potentially be of increased beneficial use for future scenarios.
My concept is to develop a hybrid centralised/swarm robot coordination system where otherwise autonomous robots can switch control over / receive objectives from a federal command centre. In this way, when the overlooking system detects a hazard, it will gain priority over the robotic ecosystem, and produce a proactive response utilising any particular robotic units it deems necessary.
With the intention of using low-cost and expendable robots, such as modifying decommissioned bomb disposal robots (in this case, substituted with a toy), fitted with a networked robot controller, the goal is to also demonstrate how a partially impaired robot could still be utilised to perform crucial tasks, with the general robotic ecosystem adapting accordingly.
Why Robots?
In many circumstances, seen from historical disasters, robotic systems were deployed to perform tasks in conditions deemed too hazardous for human personnel. The outcome of these actions displayed mixed results, from completing tasks that would otherwise be too risky to undertake, to these systems inevitably failing and resulting in human intervention anyway.
Aside from the success of some efforts, these attempts were of sound justification, with a high reward if the risk was overcome. More importantly, through using systems aimed to alleviate direct human action, these devices were still under direct human control, retaining a dependency on both human involvement, human intuition and, often, on-site presence.
Regardless of the exact design or functional factors of past or future robots, my primary ambition is focused on resolving the issue of human dependency for controlling the robots and assessing the situation. Even if a "perfect" robot could carry out the tasks, the need to be on-site to control the devices and produce a plan of action presents fatal risks that an end-to-end autonomous ecosystem may have the potential to overcome.
However, using autonomous robots, the dependency on human controllers may be resolved.
Why a Hybrid Swarm / Central Solution?
Swarm robotics provides many benefits to a range of scenarios. Autonomous robots do not necessarily require a network connection or any kind of hierarchy to formulate decisions and take action. For this reason, this decentralised approach can be beneficial for deploying robotics to map an area or search for hazards without any external dependencies or single points of failure.
However, the lack of centralisation causes its own issues. If perhaps a fire broke out, this would only be detected when a robot happens to come across it. Additionally, what happens when it is detected? If it is a large fire, a single robot may not be able to overcome it, and without a means to communicate with all the robots, only those nearby can be notified. Similarly, what if the entire group are told? Will ot be coordinated efficiently or, without a leader, will every robot do whatever it thinks best, beneficial or a hindrance?
By pairing a swarm of robots with a federal overseer, the general robot ecosystem can proactively carry out tasks autonomously, utilising the benefits of swarm robotics. However, there is also the opportunity for the central system to reprioritise objectives, inject new tasks, or even take manual control over a specific robot when the system deems it vital.
A semi-centralised approach offers benefits from taking into consideration factors a single robot would not be aware of, as well as being capable of establishing a hierarchy to enable careful management and optimising of what, how and when tasks are accomplished. Additionally, if a robot is impaired, such as visually, either from environmental factors or subsequent damage, this central manager could use information gathered from other actors and sensors within in the ecosystem to provide the robot with information that the robot has lost the ability to capture directly - The exact scenario that was demonstrated in my presentation.
Restrictions
Achieving these goals would, of course, be a monumental undertaking, if even possible. The project's scope was therefore restricted to concentrate on the primary objectives, focused on a specific scenario to produce a more realistic workload expectation for one lone worker. The chosen scenario was a "hazardous" material cleanup operation for which polystyrene substitutes were produced.
The first restriction concerned the robots. Delving into the specifics of an "ideal" robot, its design, and what it is comprised of can itself be multiple degrees and a lifetime of work. Developing a fully radiation-hardened super robot is not the intention; what is being examined is the capability of the system to effectively control a robot to complete the tasks, and that a suitable robot can be altered to connect to the system using readily available and budget-friendly controller hardware.
Therefore, the first restriction was to limit the expectations for the robots to only have the physical ability to move these polystyrene replicas to another location and support a network connection to the central computer. The robots themselves are not the target of the project, so any robot that can accommodate these needs will be considered suitable.
The second restriction was the test environment. As I do not possess an exploded nuclear reactor handy, an alternative must be made. For this project, there will be two environments to examine the system. The physical environment will be of at least a 3 x 3 grid space (though ideally more), where one grid space is the size of a robotic unit. This will be enough to scrutinise the control systems' ability to detect and locate hazards in this area and control a robot to pick up and relocate this hazard into a designated "safe disposal zone".
The second environment will be a virtual imitation of a larger workspace. This simulation will be used to test if the system can handle a multitude of simultaneous hazards and its effectiveness in controlling a small fleet of individual robots, signifying whether it can withstand a larger and more realistic scenario.
More Coming Soon!
Keep an eye out for more chapters on this project.
