Search for: All records

Autonomous Mobility-on-Demand (AMoD) systems represent an attractive alternative to existing transportation paradigms, currently challenged by urbanization and increasing travel needs. By centrally controlling a fleet of self-driving vehicles, these systems provide mobility service to customers and are currently starting to be deployed in a number of cities around the world. Current learning-based approaches for controlling AMoD systems are limited to the single-city scenario, whereby the service operator is allowed to take an unlimited amount of operational decisions within the same transportation system. However, real-world system operators can hardly afford to fully re-train AMoD controllers for every city they operate in, as this could result in a high number of poor-quality decisions during training, making the single-city strategy a potentially impractical solution. To address these limitations, we propose to formalize the multi-city AMoD problem through the lens of meta-reinforcement learning (meta-RL) and devise an actor-critic algorithm based on recurrent graph neural networks. In our approach, AMoD controllers are explicitly trained such that a small amount of experience within a new city will produce good system performance. Empirically, we show how control policies learned through meta-RL are able to achieve near-optimal performance on unseen cities by learning rapidly adaptable policies, thus making them more robust not only to novel environments, but also to distribution shifts common in real-world operations, such as special events, unexpected congestion, and dynamic pricing schemes. 

We have been interested in the development of rubisco‐based biomimetic systems for reversible CO₂capture from air. Our design of the chemical CO₂capture and release (CCR) system is informed by the understanding of the binding of the activator CO₂(^ACO₂) in rubisco (ribulose‐1,5‐bisphosphate carboxylase/oxygenase). The active site consists of the tetrapeptide sequence Lys‐Asp‐Asp‐Glu (or KDDE) and the Lys sidechain amine is responsible for the CO₂capture reaction. We are studying the structural chemistry and the thermodynamics of CO₂capture based on the tetrapeptide CH₃CO−KDDE−NH₂(“KDDE”) in aqueous solution to develop rubisco mimetic CCR systems. Here, we report the results of¹H NMR and¹³C NMR analyses of CO₂capture by butylamine and by KDDE. The carbamylation of butylamine was studied to develop the NMR method and with the protocol established, we were able to quantify the oligopeptide carbamylation at much lower concentration. We performed a pH profile in the multi equilibrium system and measured amine species and carbamic acid/carbamate species by the integration of¹H NMR signals as a function of pH in the range 8≤pH≤11. The determination of ΔG₁(R) for the reaction R−NH₂+CO₂R−NH−COOH requires the solution of a multi‐equilibrium equation system, which accounts for the dissociation constantsK₂andK₃controlling carbonate and bicarbonate concentrations, the acid dissociation constantK₄of the conjugated acid of the amine, and the acid dissociation constantK₅of the alkylcarbamic acid. We show how the multi‐equilibrium equation system can be solved with the measurements of the daughter/parent ratioX, the knowledge of the pH values, and the initial concentrations [HCO₃⁻]₀and [R‐NH₂]₀. For the reaction energies of the carbamylations of butylamine and KDDE, our best values are ΔG₁(Bu)=−1.57 kcal/mol and ΔG₁(KDDE)=−1.17 kcal/mol. Both CO₂capture reactions are modestly exergonic and thereby ensure reversibility in an energy‐efficient manner. These results validate the hypothesis that KDDE‐type oligopeptides may serve as reversible CCR systems in aqueous solution and guide designs for their improvement.