If you love sci-fi films in the ‘action’ genre, you are sure to be more than familiar with robots that appear to think, move and even respond like humans. Consider blockbusters like Terminator 2 and Wall-E and you realise that while films have taken a step forward when it comes to imagining humanoids, the scientific and AI world we live in has still not caught up to making them a reality.
Why is it difficult for autonomous robot scientists and researchers to design AI that can make robots similar – emotionally and cognitively to human beings and get them to support a multitude of tasks today that require to be done manually despite the increasingly automated world we live in? The answer to that challenge lies somewhere in the realm of embodied intelligence, or the ability to carry out meaningful interactions with the environment, and collective intelligence, or the ability to solve a task by a collective of individuals, without the single individuals having the capacity to solve such a task or even being aware of the task.
Embodied intelligence highlights the sensory impulses such as the ability to distinguish between hot and cold or simple and complex, as well as balance and other haptic characteristics that are critical for us as human beings to carry out our daily activities – walking, climbing, studying, and analysing, among others. Clearly though, we have made significant progress in this field today and are likely to see more and more advanced robots with superior embodied intelligence in the near future. Advancements in embodied intelligence can be the key enabling technology to translate abstract AIs (on phones and computers) into tangible robot-like AIs (not necessarily made of metal!).
Another dominant concern – perhaps one as old as the AI versus human intelligence debate – or older (people brought up on a diet of popular Sci-Fi shows like Star Trek will concur) is the oft-asked question of whether robots can ever have a sense of social conscience and demonstrate the same ethicality that we expect from humans in general. With this question we enter still murkier ground as there is far less research available to state conclusively if this can ever be achieved, making this issue more of a philosophical one.
A conscious robot, undoubtedly, is the need of the hour and could quite literally, usher in a world that moves to clockwork precision. While AI is often looked at with a fair degree of skepticism by naysayers that are sure it will spell the end of the human civilisation as we know it, such fears are baseless in reality as we are very far from a time when this scenario could even come to pass as it would require conscious robots, self-replicating robots, or a combination of both.
Collective intelligence is something we see often enough in nature. Colonies of ants marching along in line as they carry a large morsel of food – or a flock of birds moving in synch as they migrate to warmer climes in winter or carry raw materials to build their nests come to mind in vivid flashes. Likewise, the ability to bring together a swarm of robots capable of executing complex tasks and maneuvers – such as surveillance and monitoring of populations to ensure compliance to rules, for instance adherence to wearing masks during the pandemic, can best be carried out by a swarm.
Strictly speaking, individual ants have around 250,000 neurons and honeybees and cockroaches boast around 10,00,000! These astounding numbers still pale in comparison to human beings who possess 86 billion neurons (four times the order of magnitude). The individual in a collective is also limited when it comes to sensory impulses. In addition to their eyes, ants are equipped with antennae and are able to perceive chemicals (smell), vibrations, air currents and even tactile information. Their perception is purely local and their ability to coordinate with others in the colony happens entirely through passing on local information. Queens of social insect colonies are the exception that can, through releasing chemicals, influence the development of larvae into different classes of adult individuals, later capable of performing myriad tasks.
A swarm of robots is stronger and more resilient in different environments. Also, the intelligence of single individuals is limited when compared to an army of ants or a flock of birds or bees. Looking at the bigger picture, it is clear that robots, like these species in the natural world, can also be trained to work better in swarms. In large open spaces that require monitoring, search and rescue, these tasks are performed better through using multi-robot systems rather than single robots.
So, how do multi-robot systems function? Most of the current multi-robot systems, specifically aerial ones, rely purely on external sensing and communications. They use external infrastructure, such as GPS signals to determine their position (which require external satellites), and terrestrial infrastructures (such as Wi-Fi or 4G) for communication. With social insect societies, however, they are the infrastructure. They do not need external localisation systems, and use their own strengths, instead – ants create chains to connect a prey to a nest and to complete the collective transportation task.
Swarm robotics is an emerging field that offers infinite possibilities to study the models of control and coordination used by social insects and replicate these in highly efficient schemes for robotic control and communication. Robotic societies, working similarly to social insects, can solve the same tasks without requiring any external infrastructure. In doing so, robotic systems become infinitely scalable, and can bring together societies numbering hundreds of thousands of individuals – like locust swarms. There is merit in this study, primarily because, any infrastructure-based technology will eventually hit a limit that swarm technologies are free from.
Robotic swarm systems can be used in applications where the external infrastructure cannot be built in advance. The top use cases for these swarms in our prevailing societies is extra-terrestrial planet exploration, search and rescue following a disaster, bringing down existing built infrastructure, and underwater operations, to name a few.
Most of today’s research in swarm robotics is performed using simulation software and in controlled laboratory conditions. This is because the hardware platforms used in swarms do not yet have the “embodiment” features necessary to interact with the vastness of real-world conditions. Despite these limitations, swarm robotics offers prospects that well justify the investment they require today.
