EMERGE will implement a clear research-to-technology pathway to surpass limitations and barriers of the current state-of-the-art multi-agent collaborative systems, with potential to produce breakthroughs and open new markets in the next generation of robotic systems.
In the context of industry 4.0, collaborative robots combined with lifelong and edge learning will have the capability to go beyond the usual structured human-robot collaboration scenarios. They will be able to act in loosely structured or completely unstructured conditions—meaning that they will be able to adjust their performance according to local conditions. This will allow to decrease costs of deployment and maintenance, as well as increase robustness and interoperability of the system.
The Consortium will focus on three use cases. The first use case is modular soft robots – self‐assembling, repairing or replicating robots made from soft materials which offer higher freedom of movement, even in confined spaces, and better manipulation of delicate objects. In these robots, the body formed by a physically distributed collective needs to self-organise to account for the dynamic addition of components. The second use-case are robotic swarms – groups with a large number of robots whose behaviour arises from the interactions between themselves and with their environment. This is an example of a large-scale minimal collective where agents need coordination to achieve a collaborative goal. Finally, the third use case are collaborative robots, or cobots – robots interacting in direct contact with, or in close proximity to, humans. These represent a closer-to-market use case where interoperability is currently a significant barrier.
While robotics provides the perfect testing ground for this new framework, EMERGE also envisions impact in areas such as Internet-of-Things (IoT), smart cities and transportation, microservice-based information and communications technology (ICT) systems, and biomedical nanodevices, among others.
Use Case 1: Soft Robots
Modular soft robots are self‐assembling, repairing or replicating robots made from soft materials which offer higher freedom of movement, even in confined spaces, and better manipulation of delicate objects. The body of these robots is formed by a physically distributed collective which needs to self-organise to account for the dynamic addition of components.
EMERGE will investigate the capability of a robotic entity made of modular parts to simultaneously acquire awareness of its own body and the task to be accomplished with minimal explicit coordination from a central brain. Modular soft robots are particularly interesting because they are physically distributed mechanical systems. Therefore, each infinitesimal element of the mechanical continuum serves as an atomic unit of awareness, and basic forms of awareness can already arise from a single soft tentacle.
As key validation, we will target a local minimal collective under the form of a modular soft squid-like robot entirely made of silicon, prepared by TU Delft. Its main body will contain some central intelligence, where any number of soft tentacles can be added in a modular fashion (along with their local intelligence). Minimal or no explicit information is shared between the various parts, which will autonomously learn to swim in an unknown environment. This demonstrator will both investigate the capability of physically distributed systems to create awareness and test awareness in a scenario of contained complexity.
Use Case 2: Robot Swarms
Robotic swarms are groups of a large number of robots whose behaviour arises from the interactions among themselves and with their environment. This is an example of a large-scale minimal collective where agents need coordination to achieve a collaborative goal.
EMERGE will leverage the Swarm Testbed for Intralogistics at the University of Bristol to explore unit awareness and emergent collaborative awareness on a swarm of 20 robots called DOTS, built for distributed organisation and transport systems in intralogistics. DOTS are fast, have long autonomy (6h), can sense their environment (4 cameras, 16 laser TOF sensors), and have fast on-board processing for edge AI. Additionally, the robots can communicate through WiFi or Bluetooth and lift payloads (boxes).
The scenario is based on an autonomous warehouse with users taking robots out-of-the-box and needing the collective to gain collaborative awareness in order to achieve the task and communicate their awareness to the user. Swarm members will be artificially made minimal, their capabilities varied by evolutionary learning to understand the impact of complexity and heterogeneity on the emergence of collaborative awareness.
Use Case 3: Cobots
Collaborative robots, or cobots, are robots interacting in direct contact with, or in close proximity to, humans. These represent a closer-to-market use case where interoperability is currently a significant barrier.
EMERGE will test advantages of aware systems in standard industrial tasks involving the collaboration of more agents. We will assume no direct exchange of information between the robotic systems beyond physical medium and sensor inputs (vision, touch). The goal is to see if aware systems can autonomously understand their role in executing tasks and assess their capabilities. This implicit coordination is how humans cooperate in teams, and if achieved with artificial beings, should foster human-cobots cooperation in the long term.
Cobots have potential for large-scale industrial use and penetration in SMEs, they allow cooperation with human workers and easy restructuring of the production line. However, cobots still suffer from 2 significant issues which substantially slow down their diffusion: (i) limited individual intelligence and inability to autonomously adapt to situations and, (ii) lack of interoperability between systems by different providers. This use case will investigate if emergent awareness can solve these two challenges by putting collaborative robots in a realistic retail environment, available at TU Delft. Cobots from different companies (PAL, Boston Dynamics, Franka) will interact without direct information transfer, assessing whether collaborative awareness can emerge to a sufficient extent to perform standard tasks like filling shelves with products.