Rapid advancements in analytics, AI and self-learning algorithms are transforming cyber-physical systems (CPS), in general, and robotics, in particular. New CPS of intelligence are capable of autonomous decision-making, communicating with other machines or “assets” connected on the digital thread, and communicating with humans in the ecosystem, all in real time.
These systems consist of physical components (sensors and actuators, including mechatronic, electrical, hydraulic, pneumatic systems, etc.), software components (they host algorithms for making intelligent decisions based on the physical input received by the system), and connectivity (to link every connected asset with the other components for generating and exchanging useful data over the network).
Armed with the ability to function autonomously and take decisions based on the insights gathered from their interactions with the physical environment as well as with each other, these systems are expected to bring a transformative change in the quality of our lives in the future and form the basis of new and emerging smart infrastructure, products, and services.
Swarm intelligence, applied to robotics, is an emerging field of AI inspired by the behavioural models of social insects (ants, bees, wasps). A swarm combines the power of many minds into one, allowing the system to be smarter, more insightful, and more creative. This trait is transforming robotics, enabling physical robots to achieve a desired collective behaviour based on inter-robot interactions as well as their interaction with the environment.
The key characteristics of the swarm include autonomy, flexibility, cooperation, scalability and decentralized control. In other words, all autonomous robots work towards a common goal; and each one of the robots is autonomous/independent, but works for all/each other. For example, a swarm of bots can be released in the bloodstream to diagnose cancer and can be reprogrammed for targeted drug delivery when anomalies are detected.
Devices that constitute a swarm are typically inexpensive and possess only some of the communication and computation capabilities needed to operate as part of a collective whole. The swarm is powered by individual bots capable of operating in a fully autonomous mode and having capabilities such as self-deployment, self-repair, and self-optimization. Operating in a swarm enables the optimal distribution of computational and storage workload, and reduces communication dependencies with base systems.
Areas of application
Flexible and self-organized factories are becoming a reality as real-time data combine with human-like decision-making made possible by advances in machine- and deep-learning AI algorithms. By virtue of their autonomous decision-making and machine–to-machine communication capabilities, CPS are enabling factories to become smarter. Critical assets can be monitored remotely using data generated by plant sensors. Low latency swarm robots located near the monitored machine can quickly respond to anomalies and ensure safe operations.
The swarm behaviour of drones can be a game-changer for military and surveillance. They can be used for intelligence gathering, targeted missile attacks, and enhanced decision-making. Recently, the U.S. Department of Defense successfully carried out a demonstration of miniature drones.
These swarms can also carry out tasks deemed impossible now, primarily due to the enormity of resources required. In case of a natural disaster or calamity, drone swarms can spread out over a large area to gather real-time information about the source of calamity, analyse the impact, and locate the affected population and area. They can also be used to create a managed transportation system by coordinating various transportation modes and tracking the movement of people and vehicles.
Technological Considerations
Such unique opportunities also bring some unique challenges to the deployment of swarm robotics. To successfully implement these systems, maximise their benefits and avoid pitfalls, organizations should consider the following:
- Step-by-step adaptation: A big bang approach to migrating the existing asset ecosystem at one go can be overwhelming due to its complexity, cost and operational risks. To overcome this, a systematic approach should be adopted by identifying a business case that will have the highest potential for improvement, followed by planning a pilot project, starting with less complex implementations.
- Designing for human interventions and overrides: While designing a swarm robotics system, it is important to build in human interventions and overrides. This will help in situations where the self-controlled environments do not function as planned. Having human interventions and overrides will also enhance the system’s trustworthiness.
- Modelling, simulation and piloting discipline: When organizations complete planning and design, it is advisable to model and simulate the functioning of these systems by running pilots in controlled environments. As these systems become more autonomous and work in swarms, any challenges pertaining to the collaborative aspects of different systems need to be identified early on. This will reduce the risks and uncertainties during actual deployments, including the major risk of downtime.
- Ensuring security: As machines and processes become more connected, they will inevitably be exposed to vulnerabilities and pose significant threat to critical industrial infrastructure. Hence, deployments must guarantee human safety by establishing privacy controls in compliance with the corresponding regulations.
Organisations planning to move forward with swarm robotics implementations will need to ensure that their programmes consider all key parameters with special emphasis on security. This will help them deploy and leverage these systems as intelligent ecosystems of machines that coexist with and enhance human capabilities.
