This paper analyzes how collaborative robots can contribute to achieving some of the United Nations' Sustainability Development Goals and reflects on the advantages that a meta-design approach could bring to robot deployment in real settings. The paper highlights how true sustainability not only depends on technological innovation but also on considerations that pertain to the social sphere of the intervention, like the specific domain, the workplace, and the user community, which require infrastructures for customization, sharing, and collaboration.
The term collaborative robots, or cobots, refers to robots designed to work alongside humans in a shared
workspace, unlike traditional industrial robots that operate independently or within restricted areas
[
Collaborative robotics plays a crucial role in advancing several Sustainable Development Goals
(SDGs) [
CEUR
ceur-ws.org
address this problem, the scientific literature proposes several approaches to end-user programming
(EUP) in the robotics field [
In their survey on end-user robot programming, Ajaykumar et al. [
In the first category, visual programming languages are often proposed. They may be based on
computer-oriented notations, such as flowcharts in RoboFlow [
An alternative approach to robot program specification is the use of natural language. For instance, in [19], a web-based natural language interface is presented that supports occupational and rehabilitation therapists in defining tasks for interacting with a NAO humanoid robot. As underlined in the survey by Villani et al. [20], due to the complexity and safety-critical issues of the robot tasks, natural language programming is dificult to deploy in the manufacturing and industrial contexts.
Other approaches combine in a unique environment diferent paradigms, like the block-based interaction with natural language programming [21][22] or verbal commands with sketches that provide context and logic for the robot program [23].
Artificial Intelligence (AI) is increasingly being leveraged to simplify robot programming, making it more accessible to non-experts. Among AI-driven approaches, Large Language Models (LLMs) are emerging as powerful tools for EUP in robotics due to their ability to interpret natural language commands and generate executable code. By bridging the gap between human intent and machine execution, LLMs enable users to define robotic tasks without requiring deep programming knowledge. However, their adoption introduces critical challenges, particularly regarding code reliability, error prevention, and user validation. A key issue in LLM-based robot programming is code verification. As explored in [24], errors in LLM-generated code typically occur in the interpretation and execution phases, with a notable tendency for LLMs to overlook crucial details from user prompts. The authors propose prompt engineering techniques, such as reinforcing task constraints and structuring numerical task contexts, to reduce execution errors. Their study underscores the necessity of dedicated verification tools, including custom scripts and simulation environments, to enhance the reliability of LLM-assisted programming.
A similar concern arises in human-in-the-loop approaches where users must refine LLM-generated code. For example, the tool presented in [25] allows users to assess and modify Python or C++ code produced by ChatGPT. However, this approach assumes a certain level of programming expertise, as users must evaluate and correct the generated code. To overcome this limitation, hybrid approaches have emerged to further integrate LLMs into EUP for robotics. In [26], a system is introduced for programming pick-and-place tasks via a natural language interface. The LLM interprets user commands, translates them into structured data, and visualizes the results using Google Blockly1, enabling nontechnical users to validate and refine the robot’s task flow. This study highlights the issue of LLM non-determinism and the necessity for user oversight in verifying task execution. In [27], an LLM-driven interface assists in automating chemistry laboratory workflows. The system provides a 3D robot model alongside a chat panel where the LLM generates Python scripts for execution. However, this approach still requires users to have programming knowledge to validate correctness.
Beyond industrial and professional applications, LLMs also play a growing role in social robotics, particularly in dialogue management for human-robot interaction [28] [29]. LLMs enhance conversational lfuency by contributing to more natural and engaging human-robot interactions.
These developments illustrate the potential and challenges of AI-based EUP for collaborative robotics. While LLMs ofer a promising avenue for democratizing robot programming, issues related to reliability, interpretability, and non-determinism must be addressed. Furthermore, all the mentioned approaches to EUP in robotics are focused on facilitating coding and rarely consider other aspects related to the specific application domain, users’ profiles, or social issues that may hinder participation in end-user robot programming.
Our recent efort to improve cobot programming is represented by PRAISE (Pharmaceutical Robotic
and AI System for End Users) [30]. This system has been developed to assist pharmacists in defining
robot programs for the preparation of galenic formulations, which are medicinal preparations made in
pharmacies to meet specific patients’ needs (e.g., in cases of allergies to certain excipients or when specific
dosages are not commercially available). Such robot programs allow guiding cobots in performing
some of the steps included in the galenic production process, like mixing diferent ingredients, filling
capsules with the obtained galenic preparation, and transferring capsules into suitable containers. Unlike
adopting a traditional EUP perspective, which primarily focuses on enabling non-expert users to write
simplified code, we followed a meta-design approach. As underlined in [ 31], meta-design encompasses a
socio-technical perspective that extends beyond merely providing simplified programming tools. On the
technical side, we designed PRAISE as an EUD environment that empowers users to design, modify, and
validate robot programs through intuitive interfaces able to manipulate user-defined items. On the social
side, we observed and interviewed representatives of the target population and designed mechanisms to
foster rich ecologies of participation and user-system co-evolution [
A meta-design approach is essential because EUD systems must be usable, adaptable, and capable of supporting the diferent needs of diferent users and work environments. In the case of PRAISE, this means not only enabling pharmacists to create robot tasks but also providing mechanisms for defining new task items and sharing them within the community. The pharmacist can define the items required for a specific task or, alternatively, reuse previously created items. Furthermore, more experienced users can define and share the diferent task items with less experienced users. This shared repository facilitates collective improvements, ensuring that workflows evolve in response to individual and community-driven insights. As an example, Figure 3 shows the page used to describe the data characterizing a container: they include the Name, the Shared attribute, the Keywords (i.e., alternative names to call the same item according to users’ preference), and the definition of how the robot must reach the container to fill it with capsules. The latter can be a fixed Position acquired through robot teaching (Get position feature), or a shape obtained after processing the container photos captured through the robot camera (Get photo feature).
Following a meta-design approach in the domain of collaborative robotics requires considering the technical issues inherent to the robot itself, encompassing aspects such as programming and task execution. In addition to these intrinsic factors, the broader environmental context must be considered, including the perspectives of end users, the specific application domain, and the related context.
As reported in Section 1, collaborative robotics, on its own, can contribute to several SDGs. Adopting a meta-design approach allows one to be even more sustainable, by reinforcing the achievement of those SDGs and satisfying further SDGs. As previously mentioned, cobots may contribute to achieving SDG 3, Good Health and Well-Being, but meta-design can enhance this aspect by promoting an ecologically valid approach [32]: issues related to ergonomics, tasks, and workers’ needs can be analyzed to assign cobots the most repetitive, error-prone, and tiresome activities, leaving human workers time for high-value and creative activities, thus improving their well-being. Meta-design may also reinforce the achievement of SDG 9, Industry, Innovation, and Infrastructure by favoring cobot technology appropriation through EUD tools and infrastructures for sharing and collaboration. Similarly, achieving SDG 12, Responsible Production and Consumption, is one of the objectives of collaborative robotics: indeed, cobots are particularly suitable for small batch and customized productions. Since this kind of production often pertains to local and handcrafted work — as in our PRAISE project with pharmacists —, meta-design can play a crucial role in tailoring the technology to the specific context and workers’ needs.
As to other SDGs not mentioned before, SDG 4, Quality Education can be achieved by taking into
account the level of expertise of end users and leveraging appropriate EUD techniques and social
mechanisms to foster community knowledge development. In PRAISE, this objective is reached through
the integration of a system of shared items (objects, actions, locations, and robot tasks), enabling users
to access and reuse predefined elements created by others. This feature is particularly valuable for
novice users, who can benefit from existing user-created programs and gradually build their expertise
in robot programming without having to start from scratch. By facilitating knowledge transfer and
enabling incremental learning, PRAISE lowers the barriers to entry for pharmacists and other healthcare
professionals, ensuring that they not only use the system efectively but also progressively develop
a deeper understanding of automation and robotics principles. SDG 5, Gender Equality, and SDG 10,
Reduced Inequalities, can be pursued thanks to the customization of robotic applications to align them
with the specific competencies and literacy of the target community. By minimizing the required
knowledge to operate cobots, expertise in programming is not mandatory, which helps reduce the risk
of job displacement. As to SDG 5, it is well-known that the female percentage in STEM faculties is
significantly lower than that of males, and this is especially true for computer science curricula. The
democratization of programming and interaction with cobots can not only streamline the work of
existing employees but also create job opportunities that do not require expertise in programming or
robotics, helping to address the challenges faced by those without these qualifications. Cobots may
also contribute to SDG 11, Sustainable cities and communities, in that they can be more afordable than
industrial robots in the least developed countries and small work communities. However, their use must
be sustainable in the long run, when experts are no longer present to help customize the technology to
that specific community and needs. Meta-design, in this case, is fundamental to consider the variety of
end users, including their age, skills, and backgrounds, and create solutions favoring a rich ecology of
participation [
Figure 4 illustrates how meta-design applied to collaborative robots can contribute to achieving some of the SDGs defined by the United Nations.
In this position paper, we have explored the role that collaborative robots can play in achieving sustainability goals and how a meta-design approach to their programming and their deployment in real contexts is useful to keep a socio-technical perspective that can make the intervention really sustainable. Further aspects can be more deeply investigated to understand the impact of a meta-design approach in the considered field. For instance, the choice of a specific AI system, such as an LLM, to be integrated into the EUD environment for robot programming can become critical for sustainability, as it is essential to assess the limitations imposed by LLM providers in terms of cost, latency, transparency, and accuracy. Moreover, one must acknowledge that LLMs and current AI systems are far from being sustainable, given their substantial resource consumption, including energy and computational power, which raises concerns about their environmental impact. In summary, the long-term success and sustainability of collaborative robotics may leverage a meta-design framework that ensures flexibility, user control, and alignment with ethical and environmental considerations. This perspective reinforces the idea that technological advancements in robotics should not merely replace human labor but rather augment human expertise, fostering more eficient, adaptable, and sustainable work environments.
During the preparation of this work, the authors used ChatGPT and Grammarly in order to: grammar and spelling check, paraphrase, and reword. After using these services, the authors reviewed and edited the content as needed, thus, they take full responsibility for the publication’s content.