Meeting sustainable development goals via robotics and autonomous systems

The paper examines how robotics and autonomous systems (RAS) may influence the achievement of the UN Sustainable Development Goals (SDGs). The SDGs comprise 17 goals and 169 targets spanning poverty reduction, health, education, environmental protection, and institutions. Progress toward the SDGs has been uneven and was further hindered by the COVID-19 pandemic. Technological advances, notably RAS—systems that can sense, analyze, and act with minimal human intervention—are rapidly diffusing across sectors and are projected to be adopted by a majority of companies by 2025. Digital technologies can both enable and inhibit SDG targets. Prior work has highlighted AI’s mixed impacts across SDGs and RAS applications in specific domains (e.g., surgery, nursing, agriculture, conservation), as well as concerns regarding employment, pollution, biodiversity, and emissions. However, there has been no systematic, cross-SDG assessment of RAS. This study addresses that gap by conducting a global horizon scan to identify key opportunities, threats, and the net impact of RAS on all SDGs, and to explore potential co-benefits and trade-offs.

The authors synthesize literature indicating that AI can enable many SDG targets while potentially impeding others, particularly those tied to poverty, education, and inequalities. Prior studies document RAS benefits such as enhanced surgical outcomes, integrated nursing care, precision agriculture (e.g., robotic weed control), and biodiversity conservation (e.g., invasive species management). At the same time, concerns include labor market disruption, pollution and waste from technology lifecycles, risks of substituting ecological functions (e.g., artificial pollinators potentially displacing natural pollinators), and increased transport emissions from widespread automation. Plans to meet SDGs seldom account for RAS, and RAS development often proceeds without explicit consideration of SDGs, motivating a systematic assessment across all goals.

The study used a three-step expert-based horizon scan. Step one was an online questionnaire where 102 participants (from 23 countries; 57% high-income, 43% low- and middle-income; 36% female) each selected up to three SDGs aligned with their expertise and evaluated RAS impacts on every target within those SDGs using separate 5-point Likert scales for positive and negative impacts (with a “do not know” option). Participants provided 1913 qualitative statements describing perceived opportunities (for positive impacts) and threats (for negative impacts). They also rated the overall impact of RAS on the SDG (only positive; positive with negative; negative with positive; only negative; no impact) and judged the difficulty of predicting impacts (very hard to very easy). Ethical approval was obtained; data were anonymized. Step two grouped 44 participants into 17 SDG-focused groups (each including at least one engineer) to synthesize the step-one findings for their SDGs, identify key opportunities/threats, highlight the three targets most impacted (positively and negatively), and assess overall impact; low responses for SDG1 were supplemented by additional individual inputs; SDG2 and SDG14 groups did not contribute, yielding 15 group syntheses. Step three was an online workshop where group representatives presented syntheses and all 44 participants discussed cross-SDG co-benefits and trade-offs of RAS deployment. Qualitative data across steps were analyzed inductively to derive themes of opportunities, threats, co-benefits, and trade-offs. The prevalence of each theme was quantified as the percentage of participants mentioning it to indicate salience (not statistical validity). Likert score visualizations were generated in R (likert package).

• Five key opportunities for SDG delivery via RAS:
  • Replacing human activities: 58% of participants noted RAS can take over dangerous, repetitive, or hard-to-staff tasks across sectors (e.g., agriculture, fisheries, processing/packaging, waste/environmental management, invasive species eradication, healthcare cleaning, lab work, manufacturing, construction, infrastructure maintenance including water systems). Anticipated benefits include increased productivity, more frequent/smaller repairs, and reduced resource use.
  • Supporting human activities: 31% highlighted assistive RAS to alleviate workloads (e.g., elderly care), enhance surgical care and patient movement, facilitate sensitive health screening, and personalize education and vocational training; socially assistive robots (e.g., Nao) can enhance inclusion for people with impairments or vulnerabilities.
  • Fostering innovation: 28% saw RAS as accelerating R&D (notably in drug/vaccine development and renewables), enabling entrepreneurship, creating high-skilled jobs, and potentially reducing global inequalities via technology transfer.
  • Enhancing access: 46% indicated RAS (e.g., autonomous transport, drones) improve access to remote/dangerous areas, aid disaster relief (ambulance services), deliver medical supplies (blood, vaccines), support remote diagnosis/education, facilitate environmental research in inaccessible locations, and manage urban vertical farms/green infrastructure; autonomous vehicles could increase safety and reshape urban planning.
  • Monitoring for decision-making: 78% cited automated monitoring across infrastructure, resource distribution, wildlife, water quality, financial markets, and illegal fishing. Benefits include faster, more responsive, transparent data collection enabling assisted decision-making and broader public participation. However, monitoring alone does not ensure action.
• Four key threats that could impede SDGs:
  • Reinforcing inequalities: 51% warned RAS may exacerbate wealth concentration, interact with cultural perceptions, reduce low-skilled jobs, and amplify biases if AI is trained on biased data. Mitigations include bias-aware algorithms and diversifying the RAS workforce.
  • Negative environmental impact: 20% cited lifecycle impacts (energy use, resource extraction, disposal/pollution) and biodiversity/ecosystem disturbances (e.g., landscape simplification from sensor-based agriculture, drone disturbances to wildlife, potential deep-sea impacts).
  • Resource diversion from tried-and-tested solutions: 27% feared investment in high-tech RAS could divert resources from effective socio-political interventions (e.g., vaccination, education, emergency services), or be ineffective without social context (illustrated by misused public technologies).
  • Inadequate governance: 27% flagged uncertain regulation and data ownership, risks to privacy and human rights, potential for macroeconomic inequality if IP is concentrated, and environmental risks if automation bolsters large-scale ecosystem-degrading practices. Robust, adaptive international governance and equitable IP sharing were recommended.
• Net impact and uncertainty:
  • Overall, participants judged RAS impacts on SDGs as overwhelmingly positive; no SDG was predominantly negatively impacted; seven SDGs had >75% believing only positive impacts, with the remainder requiring managed trade-offs.
  • Difficulty of predicting impacts varied; certainty was higher for SDG9 (industry/innovation/infrastructure), SDG11 (cities), and SDG13 (climate). Figure 4 reports: Only positive impact 100%; difficulty ratings: Very easy 44%, Easy 32%, Neither 14%, Hard 14%, Very hard 0%.
• Co-benefits and trade-offs:
  • Co-benefits include links between land decontamination (SDG15) and waste management (SDG12); reducing food waste via monitoring (SDG12/SDG2); gender equity gains (SDG5) from replacing unsafe tasks in agriculture; transparency across value chains reducing modern slavery (SDG9/SDG8/SDG16); education improvements (SDG4) advancing gender, poverty, and health goals; benefits in food (SDG2) and health (SDG3) contributing to poverty reduction (SDG1).
  • Trade-offs include rebound effects from efficiency gains increasing consumption and environmental pressures; potential wealth concentration and transboundary resource inequities (e.g., water management); enhanced access/monitoring may facilitate overexploitation; inherent tensions within SDGs (e.g., SDG12 and SDG8) and risks of excluding citizens from decision-making via automated planning systems.

The horizon scan indicates RAS are poised to significantly influence SDG delivery by replacing/supporting human labor, accelerating innovation, improving access, and expanding monitoring. These mechanisms offer broad co-benefits across environmental and socio-economic goals. However, without careful governance and socio-technical integration, RAS could exacerbate inequalities, cause environmental harms, divert resources from proven interventions, and undermine inclusivity and privacy. The findings underscore the need for cross-disciplinary collaboration (engineering, natural, and social sciences) to design RAS that align with social goals and to anticipate trade-offs such as rebound effects and market concentration. Governance structures must ensure equitable access, fair data stewardship, inclusive participation in decision-making, and adaptive regulation that keeps pace with technological change. Integrating RAS considerations in SDG policy processes can help leverage benefits while mitigating risks.

RAS will increasingly shape progress toward the SDGs. This study identifies salient opportunities (task substitution/support, innovation, access, monitoring) and threats (inequality, environmental impacts, resource diversion, governance gaps), with an overall positive but context-dependent net impact. To maximize benefits and reduce harms, the authors recommend: explicitly aligning RAS design/deployment with the full SDG agenda; early, continuous multi-stakeholder engagement to set realistic expectations and co-design solutions; strengthening education and institutional capacity to improve equitable access and governance; and developing iterative, adaptive regulatory frameworks and equitable IP sharing. Future research should deepen interdisciplinary evaluation of RAS impacts on socially oriented SDGs (e.g., poverty, inequality, institutions), quantify rebound effects, assess real-world bias mitigation, and pilot governance models. Including RAS explicitly in future SDG iterations will help avoid unintended consequences and capture opportunities.

The study relies on expert elicitation rather than empirical impact evaluations; percentages reflect salience, not causal effects. Participant composition may introduce biases (57% from high-income countries; 36% female; majority with SDG rather than RAS specialization), and some SDGs had limited group synthesis (e.g., SDG2 and SDG14 did not contribute in step two; low initial responses for SDG1). Predicting future RAS impacts is inherently uncertain, especially for socially focused SDGs (e.g., SDG5, SDG10). Horizon scans identify emerging issues but do not substitute for detailed, context-specific assessments. Additionally, qualitative datasets are not publicly accessible due to privacy constraints.