The Sustainability Gap in Robotics: A Large-Scale Survey of Sustainability Awareness in 50,000 Research Articles

1. Problem Statement

The robotics field is experiencing exponential growth, driven by advancements in AI (LLMs) and hardware, transitioning from industrial automation to general-purpose and humanoid systems. However, this rapid expansion occurs against a backdrop of a global ecological crisis where humanity has already exceeded seven of nine planetary boundaries.

The core problem identified is a disconnect between the field's potential impact and its stated intent. While robotics technologies inherently possess the capacity to address global challenges (e.g., climate action, healthcare, resource efficiency), the research community largely fails to explicitly frame its work within the context of sustainability. The authors define two specific gaps:

• The Awareness Gap: The lack of explicit communication regarding social, ecological, and sustainability impacts in research papers.
• The Motivation Gap: The lack of explicit motivation by sustainability goals (specifically the UN Sustainable Development Goals or SDGs) in the research narrative.

2. Methodology

To quantify these gaps, the authors conducted a large-scale systematic survey of the robotics research landscape.

• Dataset: The study analyzed 49,751 open-access research articles from the arXiv cs.RO (Robotics) category, published between 2015 and early 2026. This period aligns with the adoption of the UN 2030 Agenda.
• Computational Pipeline: Due to the intractable volume of data for manual annotation, the authors developed a high-throughput pipeline using the DeepSeek-V3 Large Language Model (LLM).
  • Approach: A zero-shot classification method was employed. The LLM was provided with a comprehensive system prompt containing the full text of the 17 UN SDGs and their 169 targets, along with proposals from the International Federation of Robotics (IFR) on how robots can support these goals.
  • Classification Criteria: For each paper, the model was tasked with:
    1. Identifying the presence of impact keywords (Social, Ecological, Sustainability, UN SDGs).
    2. Classifying the paper as SDG Aligned (technically relevant to an SDG) or SDG Motivated (explicitly driven by an SDG goal).
    3. Providing textual evidence (quotes) and reasoning for every classification to minimize hallucinations.
• Validation & Robustness:
  • Cross-Model Benchmarking: A subset of 500 papers was re-analyzed using three other LLMs (Qwen3-32B, Llama-3.1-70B, gpt-oss-20b) to ensure trends were not artifacts of a single model's bias.
  • Consistency Checks: The model was run 200 times on two representative papers (one technical, one sustainability-driven) to verify classification stability.
  • Limitations: The authors acknowledge that the study measures scientific communication and stated intent, not actual real-world environmental impact. They also note the inherent limitations of zero-shot classification compared to fine-tuned models, though they argue the scale (50k papers) provides statistical robustness.

3. Key Contributions

1. First Large-Scale Quantitative Survey: This is the first study to systematically analyze the sustainability orientation of nearly 50,000 robotics papers, moving beyond small-scale manual assessments.
2. Definition of the "Sustainability Gap": The authors formally define and quantify the disconnect between the field's technical potential and its explicit sustainability framing.
3. Open-Source Infrastructure: The authors released the complete dataset, classification code, and an interactive Paper Analysis Tool on Hugging Face, enabling researchers to audit their own work.
4. Methodological Framework: A reproducible pipeline for using LLMs to assess scientific alignment with global policy frameworks (SDGs) at scale.

4. Key Results

The analysis reveals a stark "Sustainability Gap" characterized by low awareness and low motivation:

• Exponential Growth vs. Stagnant Awareness: Robotics research volume has doubled approximately every three years (from ~80 papers in Q1 2015 to ~3,100 in Q1 2026). However, the proportion of papers mentioning sustainability impacts has remained stagnant at <7.5% over the last decade.
• The Awareness Gap:
  • Social Impact: Mentioned in ~6% of papers.
  • Ecological/Sustainability Impact: Mentioned in <1.5% of papers.
  • UN SDG References: Explicitly mentioned in only 0.07% (33 papers) of the corpus.
• The Motivation Gap:
  • Only ~4.3% of papers are explicitly motivated by sustainability goals.
  • 95.7% of papers are published without any explicit sustainability framing.
• SDG Alignment vs. Motivation Disparity:
  • High Potential: A vast majority of papers are technically aligned with SDGs. 87.5% align with SDG 9 (Industry, Innovation, and Infrastructure), and 43.5% with SDG 11 (Sustainable Cities).
  • Low Intent: Despite this alignment, explicit motivation is extremely low in these dominant sectors. For SDG 9, less than 2% of aligned papers are explicitly motivated by it.
  • Sector Variance: Domains with direct environmental relevance (e.g., SDG 13 Climate Action, SDG 15 Life on Land) show higher motivation rates (>10-15%), but these represent a tiny fraction of the total corpus (<3% each).
• Temporal Trends: While the absolute number of sustainability-motivated papers is rising, it is purely proportional to the field's overall growth. The percentage of motivated papers remains flat, indicating no cultural shift toward intentional sustainability.

5. Significance and Recommendations

Significance:
The study provides empirical evidence that the robotics community is currently operating on a paradigm of "blind progress" or "accidental sustainability." While the technology could solve global crises, the research narrative is not steering it that way. The field risks becoming part of the problem (resource depletion, waste, inequality) rather than the solution if this trajectory continues.

Recommendations for Closing the Gap:
The authors propose a roadmap for researchers, institutions, and conferences:

1. Formalize Impact Statements: Mandate dedicated sections in papers for "Broader Impacts" and explicit SDG alignment.
2. Open-Source Research: Promote open-source code and data to democratize access (SDG 17) and reduce redundant computational waste.
3. Diversify Research: Encourage exploration in underrepresented SDGs (Climate Action, Marine Life, Gender Equality) rather than the saturated industrial/urban sectors.
4. Institutional Reform: Major conferences (e.g., ICRA, IROS) should adopt sustainability as a formal review criterion, require impact statements, and establish sustainability-focused tracks.

Conclusion:
The paper concludes that for robotics to contribute meaningfully to the 2030 Agenda, the community must transition from accidental alignment to intentional design, embedding sustainability into the foundational motivation of scientific inquiry.