Recent scholarship has increasingly conceptualized innovation ecosystems not merely as collections of firms but as relational systems involving heterogeneous actors, digital infrastructures, institutional arrangements, and territorial linkages that collectively shape innovation capacity [1]. Within these ecosystems, orchestrated interactions among knowledge producers, intermediaries, firms, users, and public institutions are essential for translating inputs into innovation outcomes [2]. Studies further emphasize that innovation ecosystems are adaptive and evolutionary, with performance contingent on actors’ abilities to coordinate resources, respond to uncertainty, and convert interactions into technological and commercial results [3][4].

Building upon this ecosystem perspective, innovation corridors extend the concept into spatial planning and regional development. A corridor functions as a coordination mechanism linking multiple innovation nodes, facilitating cross-place collaboration, and enabling circulation of knowledge, talent, and capital [5]. The effectiveness of such corridors is determined not only by the scale of resources they encompass but also by the internal relational structures that enable communication, collaboration, and knowledge conversion. Hence, the spatial layout of an innovation corridor actively shapes the opportunities for interaction and the efficiency of knowledge transfer [6]. Recent empirical studies highlight that corridors with dense, trust-rich, and well-networked nodes demonstrate higher innovation performance, suggesting that spatial proximity alone is insufficient without relational and institutional support [7].

Knowledge transfer is the process through which one actor’s knowledge is transmitted, interpreted, absorbed, and applied by another actor [8]. It encompasses explicit knowledge, which can be codified in patents, documents, or technical manuals, and tacit knowledge, which depends on repeated interactions, trust, and organizational learning [9]. The absorptive capacity of the receiving actor—the ability to recognize, assimilate, and exploit external knowledge—is critical in determining whether knowledge transfer produces effective innovation outcomes [10][11].

Spatial proximity facilitates knowledge transfer by lowering interaction costs and increasing the likelihood of contact, particularly for tacit knowledge [12]. However, proximity is necessary but not sufficient: actors may be geographically close yet organizationally disconnected, or trust may be lacking to support the exchange of valuable knowledge [13][14]. Therefore, the efficiency of knowledge transfer in spatial innovation ecosystems depends on the interplay of proximity, absorptive capacity, trust, network embeddedness, and institutional support. Weakness in any of these factors can create bottlenecks, constraining innovation despite high geographic or digital connectivity [15][16]. Empirical studies further show that digital platforms and data-driven coordination mechanisms can amplify absorptive capacity and enable knowledge spillovers across regional innovation nodes [17][18].

Overall, the literature suggests that innovation performance in spatially organized ecosystems is a function of both spatial and relational factors. While early studies focused heavily on geographic clustering, recent work emphasizes the importance of actor coordination, absorptive capacity, and institutional facilitation in enabling knowledge to flow effectively across nodes [19][20]. This theoretical insight underpins the analytical framework of this study: understanding innovation corridors requires examining not only physical connectivity but also the capacity of actors and institutions to recognize, assimilate, and utilize knowledge within spatially structured networks.

This article constructs an analytical framework linking spatial structure, actor interaction, knowledge transfer and innovation output. The Chengxi Sci-Tech Innovation Corridor is conceptualized as a spatial innovation ecosystem composed of heterogeneous actors embedded in differentiated subspaces. Spatial proximity affects the probability of interaction and collaboration; actor capabilities determine whether contact can become effective knowledge exchange; trust and network ties shape the stability and continuity of collaboration; and policy support influences talent attraction, platform provision and commercialization conditions.

Within this framework, knowledge transfer follows a dynamic chain: spatial organization creates interaction opportunities; agents select partners and form collaborative ties; knowledge is exchanged and absorbed through repeated interaction; successful transfer increases the corridor’s knowledge stock; and accumulated knowledge contributes to innovation output through commercialization and industrial application. Computer simulation is used because this process is nonlinear, cumulative and difficult to observe directly through static indicators alone.

The study adopts a hybrid model combining Agent-Based Modeling and System Dynamics. ABM is used to represent decentralized interaction among heterogeneous innovation actors, while SD is used to represent cumulative system-level feedback. This combination is appropriate because knowledge transfer in an innovation corridor emerges from both micro-level collaboration and macro-level accumulation. Individual actors search for partners, build trust, exchange know-ledge and learn from one another. Over time, these micro-level actions reshape the corridor’s knowledge stock, talent pool, network cohesion and innovation output.

Source: Constructed by the author.

Figure 3.Overall modeling framework of the hybrid ABM-SD simulation.

The ABM module includes six types of agents: universities, research institutes, established firms, startups, intermediary platforms and government actors. Each agent has attributes including spatial location, knowledge stock, absorptive capacity, trust propensity, collaboration willingness and network position. Interaction among agents follows a sequence of partner search, matching, collaboration, knowledge exchange, absorption and possible innovation output. The probability of successful transfer depends on spatial proximity, capability compatibility, trust, prior ties and intermediary support.

The SD module captures the macro-level evolution of policy support, talent pool, corridor knowledge stock, network cohesion and innovation output. Positive feedback occurs when higher policy support attracts talent and strengthens platforms, which then improves knowledge transfer conditions and increases innovation output. Balancing feedback may also occur because coordination cost, commercialization thresholds and absorptive limits prevent knowledge transfer from being automatically transformed into output. The hybrid model therefore connects local interactions with long-term system behavior.

Figure 3 illustrates the hybrid simulation framework combining Agent-Based Modeling (ABM) and System Dynamics (SD). The figure highlights how micro-level interactions among heterogeneous actors translate into macro-level outcomes, including knowledge stock, network cohesion, and innovation output, reflecting the dynamic feedback mechanisms of the corridor’s ecosystem.

Table 3 lists the six types of agents included in the ABM module—universities, research institutes/labs, established firms, startups, intermediary institutions, and government actors—along with their functional roles, typical initial strengths, main constraints, and key model attributes. This classification clarifies agent behavior in the knowledge transfer process.

Table 3. Agent types and core attributes.

Source: Constructed by the author.

Figure 4 visualizes the logic of agent interaction and knowledge transfer in the ABM module. It shows how agents select partners, exchange knowledge, update trust, and accumulate absorptive capacity, ultimately producing innovation output when commercialization thresholds are reached.

Source: Constructed by the author.

Figure 4.Agent interaction and knowledge transfer logic in the ABM module.

In each simulation step, agents first search for potential partners within the corridor network. The matching probability is determined by spatial proximity, capability complementarity, trust, prior collaboration experience, and intermediary support. Once a collaborative tie is formed, knowledge transfer occurs when the sender’s knowledge stock and the receiver’s absorptive capacity jointly exceed a transfer threshold. Trust increases after successful repeated collaboration and decreases when collaboration fails. Absorptive capacity increases gradually through successful learning, while knowledge stock accumulates through both internal creation and external knowledge absorption. Innovation output is generated when absorbed knowledge reaches the commercialization threshold and is supported by sufficient policy, platform, and market conditions.

Table 4 outlines the core behavioral rules governing agent interactions, including partner matching, knowledge transfer, trust updating, absorptive capacity change, knowledge accumulation, and commercialization. This provides a clear mapping between real-world dynamics and their model representation.

Table 4. Behavioral rules and update mechanisms in the ABM module.

Source: Constructed by the author.

The variable system reflects the central proposition that knowledge transfer performance is jointly shaped by spatial proximity, actor capability, network structure and policy support. Spatial proximity is represented by relative distance, corridor segment and accessibility. Actor capability includes knowledge stock, absorptive capacity and collaboration willingness. Network variables include degree centrality, bridge position, tie strength and repeated collaboration. System outcomes include transfer efficiency, network cohesion, knowledge stock, talent pool and innovation output.

Observed district indicators are not treated as direct measurements of individual actors. Instead, they are used as empirical anchors for constructing differentiated agent attributes. For example, Xihu’s research and university resources support higher initial knowledge stock for university and research agents. Yuhang’s digital economy and high-tech enterprise concentration support stronger absorptive capacity and network centrality for firms and startups. Lin’an’s manufacturing-oriented structure supports technology application and commercialization parameters. This conversion from statistical evidence to simulation structure makes the model empirically grounded while remaining computationally tractable.

As shown in the Appendix, the key simulation parameters are linked to observable corridor-level and district-level indicators, while theoretically informed assumptions are used where direct micro-level data are unavailable.

Table 5. Core variables and indicator construction.

Source: Constructed on the basis of corridor statistics, district bulletins and hybrid-modeling literature.

Table 6.Scenario setting for the simulation.

Source: Constructed by the author.

Seven scenarios are designed for comparison. The baseline scenario represents the current corridor structure and district heterogeneity. The spatial accessibility enhancement scenario reduces effective distance and increases cross-node contact. The spatial fragmentation scenario increases access barriers and tests the negative effect of weaker connectivity. The trust enhancement scenario raises the probability that repeated collaboration produces stable ties. The network bridging scenario strengthens intermediary platforms and cross-node connectors. The policy strengthening scenario increases public support for talent, platforms and commercialization. The combined optimization scenario integrates accessibility improvement, trust enhancement, network bridging and policy support.

Table 5 details the variables used in the simulation, their indicators, expected effect directions, and mapping to the model layer (ABM or SD). This provides transparency in how empirical data informs model dynamics.

Table 6 presents the seven simulation scenarios, including baseline, spatial proximity enhancement, spatial fragmentation, trust enhancement, network bridging, policy strengthening, and combined optimization. Each scenario is associated with core assumptions and specific objectives to test the effects of spatial, relational, and institutional interventions.

To make the simulation outputs comparable across different dimensions, the main outcome variables are reported as normalized indices with 2025 = 100. For each indicator, the normalized value is calculated as:

Knowledge transfer efficiency refers to the proportion of successful knowledge-transfer events among potential inter-agent interactions. Network cohesion reflects the density and stability of collaborative ties among agents, especially repeated and cross-node collaborations. Knowledge stock represents the accumulated knowledge generated internally and absorbed from external partners. Innovation output is defined as a composite index reflecting patents, technology commercialization, startup growth, and industrial application performance.

Table 7 defines key outcome indicators such as knowledge transfer efficiency, network cohesion, knowledge stock, talent pool, policy support, and innovation output. Normalization ensures comparability across dimensions with 2025 = 100 as the reference.

Table 7.Definition and normalization of main outcome indicators.

To ensure that the simulation results are consistent and reproducible, Table 8 outlines the key experimental settings used in the hybrid ABM-SD model. It specifies the simulation period, time steps, spatial units, agent types and numbers, platform, scenario settings, and output reporting conventions. These settings provide transparency for replicating the model and interpreting the results across different intervention scenarios.

Table 8. Simulation setup and reproducibility settings.

Source: Constructed by the author.

Figure 5.Main feedback loops in the system dynamics module.

The model uses representative district-actor agent groups rather than firm-level population counts. Therefore, the number of modeled agents refers to functional representative groups formed by combining the six actor types with the three corridor districts.

Figure 5 illustrates the primary feedback loops in the SD module, showing how policy support, talent pool, knowledge stock, and network cohesion interact over time to produce innovation output. Positive and balancing feedback mechanisms highlight the complex interdependencies in the corridor ecosystem.

The baseline simulation results were compared with observed development trends of the Chengxi Sci-Tech Innovation Corridor. The purpose of this comparison is not to claim exact numerical prediction, but to examine whether the simulated baseline trajectory is directionally consistent with recent empirical changes in the corridor. The observed increase in industrial value added, the expansion of the digital economy, the inflow of high-level talent, and the differentiated innovation profiles of Xihu, Yuhang, and Lin’an all support the model assumption that the corridor has strong knowledge-circulation capacity but still faces downstream commercialization constraints.

Before interpreting the simulation outcomes, it is important to validate the baseline model against observed trends in the Chengxi Sci-Tech Innovation Corridor. Table 9 compares empirical evidence with the corresponding model implications, demonstrating that the simulation captures the directional behavior of key system components, including industrial development, digital economy growth, talent accumulation, and district heterogeneity. This validation provides confidence that the model is capable of reflecting the real-world evolution of knowledge transfer and innovation performance across the corridor.

Table 9.Directional validation of baseline simulation results against observed corridor trends.

The baseline simulation shows that the Chengxi Sci-Tech Innovation Corridor already has strong endogenous capacity for knowledge circulation. From 2025 to 2035, knowledge transfer efficiency rises substantially, network cohesion grows moderately and innovation output continues to increase. This pattern is consistent with the corridor’s real-world foundation: a strong digital economy, multiple innovation platforms, high-level talent inflow and differentiated spatial functions. The result suggests that the corridor is not merely a passive container of innovation resources but already functions as a dynamic system of knowledge exchange.

However, the baseline trajectory also reveals an important structural tension. Knowledge transfer efficiency grows much faster than innovation output. This divergence is consistent with the structure of the hybrid model. In the ABM module, improved contact frequency, repeated collaboration, and stronger matching can quickly increase the number of successful knowledge-transfer events. However, in the SD module, innovation output depends not only on transferred knowledge, but also on absorptive capacity, commercialization threshold, talent accumulation, platform support, and market application conditions. Therefore, knowledge can circulate faster than it is commercialized. The baseline result suggests that the corridor’s main bottleneck is not knowledge circulation itself, but the downstream conversion of transferred knowledge into patents, products, startups, and industrial output. This means that the corridor is relatively strong in generating, circulating and recombining knowledge, but the conversion of knowledge into commercialized and industrialized outcomes remains slower. Such a gap may result from organizational learning barriers, technology maturity limits, financing constraints, intellectual-property coordination and the institutional complexity of science-to-industry transformation. Therefore, planning and governance should not only promote interaction, but also strengthen commercialization channels and downstream application capacity.

Source: Simulation results.

Figure 6.Baseline scenario trajectories from 2025 to 2035.

Figure 6 illustrates the evolution of knowledge transfer efficiency, network cohesion, knowledge stock, talent pool, policy support, and innovation output under the baseline scenario. It highlights that knowledge transfer accelerates faster than commercialization and industrial output, identifying potential bottlenecks in the corridor.

Table 10 presents the final values of outcome indicators under the baseline scenario, showing the relative growth of knowledge transfer, network cohesion, innovation output, and other system measures.

Table 10.Baseline simulation outcomes in 2035 (2025 = 100).

Source: Simulation results.

The spatial accessibility enhancement scenario improves knowledge transfer efficiency relative to the baseline. Reduced effective distance increases the probability of contact among agents in different sections of the corridor and allows knowledge resources to circulate more frequently between Xihu, Yuhang and Lin’an. Nevertheless, the effect on innovation output is smaller than the effect on transfer efficiency. This indicates that spatial proximity can create more opportunities for exchange, but cannot by itself ensure successful absorption or commercialization. Accessibility is therefore a necessary but not sufficient condition for innovation performance.

The trust enhancement scenario produces a stronger improvement in collaboration stability and innovation output. Trust reduces transaction costs, increases willingness to share higher-value knowledge and supports repeated collaboration. In innovation ecosystems where tacit knowledge and uncertain technological projects are important, trust is especially critical because actors may hesitate to disclose valuable knowledge without reliable relationships. The simulation therefore confirms that relational quality is a key mechanism linking interaction frequency to transfer effectiveness.

The network bridging scenario also improves performance by connecting otherwise separated nodes and actor groups. Intermediary platforms, incubators, proof-of-concept centers and technology transfer agencies play an important role in reducing structural holes. Their function is not simply administrative; they translate knowledge across organizational boundaries, match complementary actors and help weaker nodes access core resources. This result is particularly relevant for the Chengxi corridor because excessive concentration around central nodes may limit the diffusion of knowledge to peripheral or application-oriented spaces.

Figure 7 illustrates how enhancing spatial proximity affects the knowledge transfer efficiency across the corridor. The results show that reducing effective distance increases the probability of contact among agents in different districts, facilitating more frequent exchanges of knowledge. However, the figure also highlights that while spatial proximity improves transfer efficiency, its impact on ultimate innovation output is more limited, indicating that proximity alone is not sufficient to ensure successful knowledge commercialization.

Source: Simulation results.

Figure 7.Effects of spatial proximity on the knowledge transfer efficiency index.

Source: Simulation results.

Figure 8.Effects of trust enhancement and network bridging on innovation output.

Figure 8 presents the impact of relational trust and network bridging on innovation output. The figure demonstrates that increasing trust among agents reduces transaction costs and strengthens repeated collaborations, while network bridging enhances connectivity across otherwise disconnected nodes. Together, these relational mechanisms contribute to higher innovation performance, emphasizing the importance of trust and intermediary platforms in the knowledge transfer process.

Policy strengthening improves talent accumulation, knowledge stock and innovation output. The mechanism is cumulative: stronger public support improves platforms, attracts talent, lowers collaboration barriers and increases the capacity of agents to absorb and apply knowledge. However, policy support also has limits if it is not connected with spatial and network mechanisms. Funding and services may increase resources, but without effective cross-node linkages they may reinforce existing concentration rather than produce corridor-wide diffusion.

The combined optimization scenario produces the strongest long-term improvement among all scenarios. It simultaneously improves spatial accessibility, strengthens trust, enhances bridging ties and increases policy support. This combined effect is larger than the effect of any single intervention because knowledge transfer is a system process. Accessibility increases contact opportunities, trust improves the quality of exchange, bridging expands the reach of networks and policy support strengthens the institutional environment for absorption and commercialization. The result suggests that high-quality innovation corridor development should rely on coordinated governance rather than isolated policy tools.

The final scenario comparison also shows that spatial fragmentation produces the weakest results and can reduce innovation performance by increasing effective distance, weakening collaboration frequency and limiting cross-node knowledge diffusion. This finding is important because corridor development is vulnerable to institutional segmentation, uneven infrastructure, separated platform systems and weak communication among districts. A corridor may appear continuous on a map, but if actors do not interact effectively, the innovation ecosystem remains fragmented.

Figure 9 compares the effects of policy strengthening alone versus combined optimization across spatial, relational, and institutional dimensions. The figure shows that while policy support improves talent accumulation and knowledge stock, the combined optimization scenario produces the strongest long-term effect, illustrating the synergistic benefits of coordinated interventions across multiple mechanisms.

Table 11 presents a comparative summary of all simulation scenarios in 2035, normalized to 2025 = 100. The table highlights how different interventions—ranging from baseline, spatial proximity enhancement, trust enhancement, network bridging, policy strengthening, to combined optimization—affect key outcome indicators such as knowledge transfer efficiency, network cohesion, knowledge stock, and innovation output. This comparison provides clear evidence of the relative effectiveness of each scenario, illustrating that the combined optimization scenario achieves the highest overall performance and confirming the importance of integrating spatial, relational, and policy mechanisms.

Source: Simulation results.

Figure 9. Effects of policy intervention and combined optimization on innovation output.

Table 11.Final scenario comparison in 2035 (2025 = 100).

Source: Simulation results.

Source: Simulation results.

Figure 10.Transfer efficiency and innovation output frontier across scenarios.

Figure 10 visualizes the trade-off and frontier between knowledge transfer efficiency and innovation output under all seven scenarios. The figure highlights that the combined optimization scenario (S6) achieves the highest levels in both dimensions, whereas spatial fragmentation (S2) leads to the lowest outcomes. This reinforces the conclusion that systemic coordination among proximity, trust, network bridging, and policy support is essential for maximizing innovation performance.