2.1. Development Methodologies for Service Robots and Autonomous Product-Infrastructure Service System

Advancements in robotics are reshaping urban systems by integrating service robots into CPS and robotic autonomous systems (RAS) to enhance service delivery and efficiency (Reference Macrorie, Marvin and WhileMacrorie et al., 2021). These robots, part of smart city ecosystems, rely on digital infrastructure for communication and operation (Reference NagenborgNagenborg, 2020), mirroring past shifts like road adaptation for trams (Reference Macrorie, Marvin and WhileMacrorie et al., 2021). Current research addresses isolated aspects, such as information and communication technology’s (ICT) role in urban services (Reference Rivera, Amorim and ReisRivera et al., 2020) and fleet optimization (Reference Gupta, Kremer, Park and GöhlichGupta et al., 2022), but overlooks systemic challenges like fleet coordination, depot placement and infrastructure dependencies. Service robots in urban environments are evolving as smart products integrated with digital systems, yet design method-ologies often overlook the interconnected life cycles of robots, infrastructure, and services (VDI:2206 , 2021). Standards like ISO/IEC/IEEE 15288-5:2023 focus on iterative development but lack frameworks for managing these dependencies (ISO:15288 , 2023). Models like gPCS and modular product architect emphasize sustainability and modularity but fail to fully integrate robot and infrastructure life cycles (Reference Gräßler and PottebaumGräßler and Pottebaum, 2022; Göhlich et al., 2022). Frameworks for aligning product and service life cycle management overlook infrastructure design, critical for efficiency and scalability (Reference Gräßler and PottebaumGräßler and Pot-tebaum, 2021). Addressing interdependencies between robots, infrastructure, and operations is essential for sustainable, scalable urban robotics.

2.2. LCA for Autonomous Product-Infrastructure Service System Development

Integrating LCA into PSS development is crucial for addressing sustainability concerns (Reference Neramballi, Sakao, Willskytt and TillmanNeramballi et al., 2020). Early-stage LCAs, known as ex-ante or prospective LCAs, support sustainability-oriented design by comparing emerging technologies with established ones (Cucurachi et al.,Reference Cucurachi, van der Giesen and Guinée2018). However, this Phase faces uncertainty and data limitations (Reference Bergerson, Brandt, Cresko, Carbajales-Dale, Lean, Matthews, Coy, Manus, Miller and MorrowBergerson et al., 2020), complicating assessments (Het-herington et al., 2014). Key challenges include benchmarking against commercial products, scaling lab processes, data reliability, and managing uncertainty (Reference Hetherington, Borrion, Griffiths and McManusHetherington et al., 2014). Despite existing frameworks, prospective LCAs remain exploratory, providing insights into constraints and key impact factors (Reference Bergerson, Brandt, Cresko, Carbajales-Dale, Lean, Matthews, Coy, Manus, Miller and MorrowBergerson et al., 2020). Comparative LCAs benchmark new PSS against existing processes, requiring technology maturity alignment (Reference Hetherington, Borrion, Griffiths and McManusHetherington et al., 2014). Technology Readiness Levels (TRLs) and Manufacturing Readiness Levels (MRLs) help assess and scale emerging technologies (Reference Gavankar, Suh and KellerGavankar et al., 2015). Scale-up methods (Reference Tsoy, Steubing, van der Giesen and GuinéeTsoy et al., 2020) model scenarios and market impacts, with most research focusing on chemical and waste treatment efficiency (Reference Cucurachi, van der Giesen and GuinéeCucurachi et al., 2018). For PSS, LCA must define component relationships, manage data complexity, and evaluate system variants (Reference Marimon Viadiu, Casadesús Fa and Heras SaizarbitoriaMarimon Viadiu et al., 2006). Challenges include predicting rebound effects, behavior changes, and knowledge gaps (Reference van Loon, Diener and Harrisvan Loon et al., 2021). System-level assessments often prioritize materials and energy over reuse and recycling (Reference Chen and HuangChen and Huang, 2019). Despite uncertainties, prospective LCA offers valuable insights for PSS and emerging technologies, aiding in sustainability exploration and impact identification.

4.1. MARBLE based Autonomous Product-Infrastructure Service System

The MARBLE (Mobile Autonomous Robot for Litter Emptying) use-case, an autonomous smart service robot developed at the Technische Universität Berlin to empty street litter bins, exemplifies a smart autonomous product (Reference Göhlich, Syré, van der Schoor, Jefferies, Grahle and HeideGöhlich et al., 2022). Its autonomous product-infrastructure service system (Illus-trated in Figure 2) and possible scenarios were developed using the CLAPS methodology (Reference Gupta and GöhlichGupta and Göhlich, 2024).

Figure 2.

Smart product service system framework for use case MARBLE

Phase 2 of the CLAPS methodology revealed that the existing MARBLE prototype as a standalone product (Scenario 1) could not achieve the service goals of providing comparable or lower energy consumption and service time than those of municipal employees using diesel and electric vehicles (Scenarios 8 and 9) (Reference Bräutigam, Gupta and GÖhlichBräutigam et al., 2022). Therefore, as a potential solution, an additional assistance infrastructure in the form of a mothership was introduced. These service elements (Scenario 2) were implemented in a simulated design of the system for the SEA Monbijoupark in Berlin, involving the emptying of 52 litter bins, which led to reductions in both energy consumption and operation time (Reference Bräutigam, Gupta and GÖhlichBräutigam et al., 2022). However, as shown in Table 1, the operation time values still did not meet the desired benchmarks. Consequently, alternative solutions for achieving efficient operation in terms of both operation time and energy consumption were explored through applying Phase 4 of the CLAPS methodology, including fleet-based (Scenarios 4 and 6) operations (Gupta et al., 2022) or integrating smart infrastructure, such as smart litter-bin operations (Scenarios 3, 5 and 7) (Reference Gupta, van der Schoor, Bräutigam, Bladinieres Justo, Umland and GöhlichGupta et al., 2022). An autonomous product-infrastructure service system as smart PSS framework for MARBLE has been illustrated in Figure 2. The framework in Figure 2 illustrates how various service elements such as the MARBLE robot, its fleet, mothership, and smart litter-bins enable the service process. The overall process is managed through operation management and route planning for the autonomous robot-infrastructure service system. In the Monbijoupark SEA, with 52 litter-bins to be emptied, the physical space encompasses stakeholders, service processes, IoT devices in both the robot and litterbins, and the overall service environment. The cyber space comprises the operation management and route planning components of the service system. Table 1 lists scenarios for MARBLE’s autonomous product-infrastructure service system, detailing service elements in relation to energy consumption and operation time, compared to reference scenarios involving the municipality’s use of diesel and electric vehicles.

Table 1.

Energy consumption and operation time for different scenarios for singular shift

Based on eight PSS types (Reference TukkerTukker, 2015), this research focuses on a result-oriented autonomous product-infrastructure service system as PSS, integrating product and service design to regulate energy consumption, operation time, and CO₂-Eq emissions for the end user.

4.2. Evaluation of Scenarios with LCA

To evaluate the environmental performance of various scenarios, an LCA, following ISO 14040 standards is conducted in Phase 4 of CLAPS methodology. LCA quantifies environmental impacts across a product’s, service’s or process’s life cycle, from raw material extraction to disposal, through four iterative Phases: goal and scope definition, life cycle inventory analysis, life cycle impact assessment, and interpretation. This study compares scenarios to quantify environmental impacts and provide recommendations for reducing Global Warming Potential (GWP100) in tons CO₂-eq. The cradle-to-grave analysis includes raw material acquisition, production, maintenance, usage, and end-of-life treatment. The functional unit is emptying 52 bins over ten years in Germany. Figure 3 illustrates the processes and system boundaries. The life cycle inventory was built using the Ecoinvent database (consequential) (Ecoin-vent Centre, 2022), and the ReCiPe 2016 (Midpoint, H) method was applied for impact assessment. This method incorporates predefined characterization factors to quantify impacts. These factors translate emissions and resource use into environmental impact categories. Electricity production emissions were calculated based on 2023 data, with projections for 2030 and 2045 modeled and interpolated. The study uses the average emissions for 2026–2035. Scenarios simulate only part of daily operations, while actual vehicle and robot operation spans 16 hours per day (two 8-hour shifts). To comprehensively evaluate the MARBLE autonomous product-service system, unmodeled operating times are accounted for using two LCA approaches: allocation and system expansion. The system extension approach is traditionally used in consequential analyses. We estimate additional distances during the remaining operation time (16 hours – Operation Time for Scenarios) and calculate emissions based on distance-related energy consumption. This additional operation is incorporated into the system. For comparison, an allocation approach distributes production, maintenance, and EoL emissions across the scenarios. Using the operation time, we calculate a multiplication factor (Operation Time for Scenarios/16h) and apply it to allocate emissions fairly. Cradle-to-grave assessments for various scenarios, including production, maintenance, operation, and end-of-life (EoL) Phases, are illustrated in Figure 4, using three approaches: no allocation or system expansion (Approach B0), allocation (Approach A), and system expansion (Approach SE).

Figure 3.

Process flow for conducting LCA for MARBLE’s smart PSS

The results of the scenarios using Approach B0 show the lowest emissions for SC₁, followed by SC₂, and SC₉. The highest emissions result from SC₆ with three MARBLE, one mothership and without smart litter-bins, closely followed by SC₇ which includes smart litter-bins. In both scenarios, the production emissions contribute most to total emissions. SC₈ has the fourth lowest emissions, despite using fossil fuels for the operation. The results with allocation (Approach A) differs most from the results using Approach B0 and SE. Here, SC₉ - using the electric reference vehicle without MARBLEs - exhibits the lowest emissions. This is caused by the allocation of production, maintenance and EoL emissions, while operation emissions are the lowest for SC₉ (also Approach B0 but not SE). While the highest operation emissions result for SC₈, the diesel reference vehicle has the lowest non-operation emissions and is therefore the fifth lowest emission scenario. Both reference vehicles gain the highest reductions from allocation, caused by the minimum operation time of both scenarios leading to the lowest allocation factors. SC₅ shows the second highest and SC₁ the second lowest emissions. Approach SE shows the lowest emissions for SC₁, followed by SC₂, and SC₉ (like in Approach B0). In comparison to Approach B0, Approach SE reveals the high impact of operation emissions from fossil fuels, as SC₈ is now the worst option in terms of emissions. While for Approach A, the scenarios using smart litter-bins have lower emissions than their counterparts without smart litter-bins, for Approach SE, the scenarios using smart litter-bins show higher emissions. This is caused by the different scopes of both approaches: Approach A focuses on the defined scenarios and can therefore display the impact of smart litter-bins within the given scope. However, with system expansion, the scenarios have diverging scopes. Here, the reduced operation time through using smart litter-bins leads to increased mileage for the expanded operation. Therefore, the scenarios with smart litter-bins, using the same number of MARBLEs, can cover a greater area and more litter-bins.

Figure 4.

Carbon dioxide emissions in tons over a life cycle period of 10 years for various MARBLE and diesel as well as electric reference scenarios from municipality

A potential solution is introducing a functional unit based on the number of emptied litter-bins or served area. However, due to significant variations in litter-bin distribution and utilization across Berlin, this approach is ineffective. Scenarios SC₁, SC₂, SC₃, and SC₉ show the most promising environmental results and should be further explored. Notably, scenarios using reference vehicles require additional human labor, which is not accounted for in the LCA operation Phase. Future work will examine how human labor can be integrated when comparing autonomous and reference scenarios. We present three approaches for calculating results. Approach B0 excludes allocation and system expansion, enabling quick implementation but limiting meaningful comparison and fully depicting fleet operations. Pro-duction, maintenance, and EoL emissions are also overestimated. Approach A addresses this through allocation, distributing emissions based on operation time. While not standard in consequential LCA, it simplifies comparisons within the same scope and may be suitable for product development. System expansion, though an option, requires more effort, which must be assessed case by case. Approach SE applies system expansion, a more complex method that calculates additional operation time, distances, and energy consumption. However, it limits scenario-specific impact analysis, reducing its usefulness in product development. Nonetheless, it holds significant potential for consequential LCA, modeling emissions for entire systems and interdependencies. These insights are valuable for high-level decisions, particularly in fleet design.

4.3. Addressing Eco-design

To address eco-design principles, an evaluation is needed that assesses environmental impact alongside development and operational constraints. The evaluation of the MARBLE autonomous product-infrastructure service system focuses on Scenarios 3, 5, and 7 due to their minimal operation time and alignment with reference scenarios. Based on LCA results (Figure 4) and operational data (Table 1), this analysis adheres to eco-design principles by emphasizing sustainability and efficiency. These findings inform Phase 5 of the CLAPS methodology, guiding efforts to minimize ecological impacts while maintaining operational feasibility, supported by value-based analysis (Reference Bender and GerickeBender and Gericke, 2021). Key metrics include energy consumption (reflecting system efficiency), operation time (indicating productivity), investment costs (balancing financial and sustainability goals), and CO₂-Eq emissions (highlighting ecological performance). Using Approach A, emissions are allocated proportionally to operational time for scenario-specific evaluations. Table 2 summarizes these metrics, with CO₂-Eq emissions assigned the highest weight (40%), emphasizing the ecological focus of the paper. These weightings are assigned based on the requirements specified by external stakeholders in Phase 2 of the CLAPS methodology for the use case, with a focus on operational parameters (20% each), costs (20%), and ecological impact (40%).

Table 2.

Evaluation of chosen scenarios with value-based analysis for further development

Scenario 3 achieves a score of 2.4, excelling in energy consumption and investment costs but with slightly higher operational time and moderate emissions. Scenario 5 scores 2.8, performing well in energy efficiency and operation time but incurring higher investment costs and CO₂-Eq emissions. Scenario 7 scores 3.0, with lower operation time but higher costs and emissions, which affect its overall sustainability. While Scenario 7 offers operational benefits, its higher costs and emissions necessitate further refinement to meet sustainability goals. However, Scenario 5, with a score of 2.8, offers a balanced performance, particularly excelling in CO₂-Eq emissions reduction, which carries the highest weight (40%). Irrespective of not achieving the lowest score, Scenario 5 offers the best balance based on the quantified evaluation of energy consumption, operation time, investment costs, and CO₂-Eq emissions using Approach A, making it the recommended choice for further development. This structured quanti-fied evaluation ensures that the selection prioritizes eco-design principles while maintaining operational feasibility.