Artificial Intelligence for Climate Adaptation: Reinforcement Learning for Climate Change-Resilient Transport
Authors: Miguel Costa, Arthur Vandervoort, Carolin Schmidt, João Miranda, Morten W. Petersen, Martin Drews, Karyn Morrisey, Francisco C. Pereira
Organizations: Department of Technology, Management and Economics, Technical University of Denmark, 2800 Kgs. Lyngby, Denmark · Present address: School of Management, Technical University of Munich, 80333 Munich, Germany. · Present address: Saint-Gobain Distribution Denmark, 2605 Brøndby, Denmark. · J.E. Cairnes School of Business and Economics, University of Galway, Galway, H91 TK33, Ireland
Climate change is expected to intensify rainfall and, consequently, pluvial flooding, leading to increased disruptions in urban transportation systems over the coming decades. Designing effective adaptation strategies is challenging due to the long-term, sequential nature of infrastructure investments, deep climate uncertainty, and the complex interactions between flooding, infrastructure, and mobility impacts. In this work, we propose a novel decision-support framework using reinforcement learning (RL) for long-term flood adaptation planning. Formulated as an integrated assessment model (IAM), the framework combines rainfall projection and flood modeling, transport simulation, and quantification of direct and indirect impacts on infrastructure and mobility. Our RL-based approach learns adaptive strategies that balance investment and maintenance costs against avoided impacts. We evaluate the framework through a case study of Copenhagen's inner city over the 2024-2100 period, testing multiple adaptation options, and different belief and realized climate scenarios. Results show that the framework outperforms traditional optimization approaches by discovering coordinated spatial and temporal adaptation pathways and learning trade-offs between impact reduction and adaptation investment, yielding more resilient strategies. Overall, our results showcase the potential of reinforcement learning as a flexible decision-support tool for adaptive infrastructure planning under climate uncertainty.
Figures & tables
Figure 1: Integrated Assessment Model using reinforcement learning to find the best sequence of transport-related adaptation policies for rainfall events in Copenhagen from 2024–2100.
Exp.
Climate Scenario
BO
RL
A
RCP2.6
-119.58 ± 0.25
-118.84 ± 1.10
RCP4.5
-120.10 ± 0.38
-119.01 ± 0.85
RCP8.5
-121.24 ± 0.30
-119.23 ± 1.28
B
RCP2.6
-121.19 ± 1.45
-117.87 ± 0.57
RCP4.5
-120.51 ± 1.71
-117.09 ± 0.64
RCP8.5
-123.71 ± 1.42
-119.88 ± 1.54
Table 1: Total reward comparing the BO vs. our RL-based strategy. Average results over three runs are reported ± standard deviations for the two experiments (Exp. A and B) and the three climate scenarios (RCP2.6, 4.5, and 8.5).
Figure 2: Comparison of different adaptation strategies. Total reward (top-left, scaled by 108 ), and reward components are shown: infrastructure damage costs (top-middle), travel delays (top-right), travel cancellations (bottom-left), actions direct costs (bottom-middle), and action maintenance costs (bottom-right). All values correspond to Danish Krone (DKK). Please note the different y-axis scale across figures.
Figure 3: Average (and standard deviations shaded) for all five reward components over the period 2024–2100 under the RCP4.5 climate scenario. While Random Control reduces impacts through high and persistent investment costs, the RL policy achieves sustained impact reduction with substantially lower and more stable adaptation expenditures. All values correspond to Danish Krone (DKK). Please note the different y-axis scale across subfigures.
Figure 4: Adaptation measures taken over time per zone in Copenhagen’s city center for one episode (run) for a RCP4.5 scenario.
Figure 5: Density of actions used over time for 10 runs for a RCP4.5 scenario. Bioretention Planters are shown in top-left, Soakaways in top-right, Storage Tanks in bottom-left, and Porous Asphalt in bottom-right.
RCP2.6
RCP4.5
RCP8.5
Infrastructure damage
8.00 ± 0.18 ×108
7.90 ± 0.13 ×108
8.05 ± 0.19 ×108
Travel delays
6.55 ± 0.10 ×109
6.01 ± 0.06 ×109
5.54 ± 0.17 ×109
Travel cancellations
1.63 ± 0.10 ×109
1.61 ± 0.09 ×109
1.49 ± 0.10 ×109
Action costs
1.73 ± 0.09 ×109
2.32 ± 0.11 ×109
3.27 ± 0.34 ×109
Action maintenance costs
3.46 ± 0.32 ×107
4.23 ± 0.21 ×107
1.88 ± 0.57 ×108
Total reward
1.07 ± 0.02 ×1011
1.08 ± 0.01 ×1011
1.13 ± 0.02 ×1011
Table 2: Total reward and costs per reward component using different climate scenarios (RCP2.6, RCP4.5, and RCP8.5) over different 10 runs. Average results are reported ± standard deviations.
Belief Scenario
Reality Scenario
Reward
RCP2.6
RCP2.6
-107.42 ± 1.55
RCP4.5
-107.53 ± 1.54
RCP8.5
-109.94 ± 1.19
RCP4.5
RCP2.6
-107.42 ± 1.55
RCP4.5
-107.67 ± 1.29
RCP8.5
-109.45 ± 1.15
Table 3: Total reward using an RL strategy that was designed using various climate beliefs and reality scenario pairs. Average results are reported ± standard deviations.
Appendix figures & tables1 asset
Supplementary material from the paper’s appendix.
Appendix
Adaptation Measure
Effect
Implementation details
Implementation costs (DKK)
Maintenace costs (DKK/year)
Lifetime (years)
Bioretention planter
Captures 2 m3 of water runoff
Applied every 30m of road
14312 per planter
24 per planter
40
Soakaway
Captures 5.4 m3 of water runoff
Applied once per road
7273
1.9
30
Storage tank
Captures 15 m3 of water runoff
Applied once per road
104896
5
50
Porous asphalt
Captures 0.3m of water depth
Replaces the road with this pavement type
995.77 per m2 of surface area
0.635 per m2 of surface area
30
Pervious concrete
Captures 0.45m of water depth
Replaces the road with this pavement type
1199.81 per m2 of surface area
0.635 per m2 of surface area
30
PICP
Captures 0.25m of water depth
Replaces the road with this pavement type
1046.78 per m2 of surface area
5.195 per m2 of surface area
50
Appendix
Table 4: Adaptation measure details used in our framework.
Reinforcement learning studies how agents learn sequential decisions from rewards and interaction, balancing exploration, performance, safety, and adaptation.
In modern intelligent transportation systems (ITS), particularly in freight transportation and logistics, real-time route planning is crucial. It presents unique challenges driven by high uncertainty in service requests, where the number of service customers can vary drastically, ranging from hundreds to thousands. Existing neural methods struggle to maintain performance under such significant variations, which severely limits their practical applicability. To address this crucial shortcoming, this work proposes a novel Instance-Conditioned Adaptation Model (ICAM) designed for better large-scale generalization. In particular, we design a simple yet efficient instance-conditioned adaptation function that adjusts the policy based on the specific geometry and density of the current traffic scenario to improve model adaptability with minimal computational overhead. Furthermore, we propose a powerful yet low-complexity instance-conditioned adaptation module to generate better solutions for instances across various scales. Extensive experiments on synthetic, benchmark, and real-world instances demonstrate that ICAM can consistently achieve promising generalization performance across four widely studied large-scale route planning scenarios. Notably, our proposed method delivers high-quality solutions with remarkably fast inference speed, providing a scalable and efficient solution for real-time intelligent transportation operations. Our code is available at https://github.com/CIAM-Group/ICAM.
Changliang Zhou, Xi Lin, Zhenkun Wang +3
Guangdong Provincial Key Laboratory of Fully Actuated System Control Theory and Technology, School of Automation and Intelligent Manufacturing, Southern University of Science and Technology, Shenzhen 518055, China · School of Mathematics and Statistics, Xi’an Jiaotong University, Xi’an 710049, China · Huawei Noah’s Ark Lab, Hong Kong SAR, China +1
Dynamic traffic variations in Open Radio Access Networks (O-RAN) lead to drift, which degrades the performance of Artificial Intelligence/Machine Learning (AI/ML) models. Traditional retraining approaches maintain forecasting accuracy but incur high computational cost and may lead to violations of Service Level Agreements (SLAs). This work proposes a Q-learning-based adaptive retraining approach that formulates the retraining decision as a Markov Decision Process (MDP), where a Reinforcement Learning (RL) agent learns a policy that balances forecasting accuracy and retraining cost. The proposed approach incorporates a multi-expert Long Short-Term Memory (LSTM) ensemble to mitigate catastrophic forgetting and improve robustness across diverse traffic conditions. Experimental results show that the proposed approach effectively reduces retraining overhead compared to greedy and random baselines, while maintaining system performance within predefined limits.
Extreme rainfall can rapidly degrade road-network functionality (RNF), disrupting emergency mobility, accessibility, evacuation, logistics, and infrastructure restoration. Existing flood-transportation impact models are often difficult to deploy across regions because they rely on local calibration, intensive simulation workflows or target-region retraining. This study proposes G-R2R, a transferable spatiotemporal graph-learning framework that reframes rainfall-induced road-network disruption as a region-conditioned, topology-flexible graph forecasting problem for unseen transportation networks. The framework integrates ego-centered subgraph sampling, terrain-conditioned graph representation, region-conditioned dual-level static encoding, and dynamic rainfall-network feature learning. Static geographic and infrastructure attributes are fused with rainfall forcing and recent RNF states to forecast node-level RNF across heterogeneous road networks. G-R2R is evaluated using simulation-derived RNF sequences from ten flood-prone regions in Wenzhou, China, under unseen-region testing protocols. It outperforms representative forecasting baselines, achieving a stabilized MAPE of 2.33% and R2 of 0.855 under the fixed unseen-region benchmark. Leave-one-area-out evaluation yields stabilized MAPE values of 2.30%-3.54%. Ablation analysis supports the contributions of the terrain-conditioned graph representation and dual-level static encoding. A historical-event replay of Typhoon Fitow further shows that G-R2R reproduces principal temporal trajectories and node-level spatial patterns of the simulation-derived RNF reference under observed rainfall forcing without target-region retraining. These results support rapid and transferable node-level RNF forecasting across the regional road-network domains represented in this study.
Junyan Wang, Naiyu Wang
College of Civil Engineering and Architecture, Zhejiang University, Hangzhou 310058, China