Search

In the rapidly transforming landscape of urban transportation, Mobility-as-a-Service (MaaS) has emerged as a potentially transformative solution to adapt to fluctuating travel demand. However, despite arguments in the literature suggesting that uncertainty in travel plans may influence travelers’ preferences for MaaS, the extent and mechanisms of this impact remain relatively underexplored. This study aims to bridge this gap by evaluating travelers’ preferences toward MaaS subscriptions, mode choices, and their reactions to uncertainty within a risky-choice modeling framework. We incorporate a linear-specified benchmark model and two nonlinear models (one based on Expected Utility Theory with Constant Absolute Risk Aversion adjustments, and the other on Prospect Theory), to estimate travelers’ risk perceptions and their valuation of gains and losses associated with travel time and costs caused by occasional activities during commute trips. The data utilized in our study originates from a specifically designed stated preference survey carried out in Hong Kong. Our findings present insightful observations into MaaS bundle configurations, socio-demographic patterns, and notably, behaviors under uncertain circumstances. In the face of uncertainty, we observe a noticeable increase in reluctance to adopt MaaS, and a heightened preference for taxis. The models also underscore significant risk aversion and diminishing sensitivity concerning travel time and cost, with these tendencies proving more pronounced for travel time compared to cost. These insights are valuable for policymakers and MaaS operators, offering an enhanced understanding of users’ risk attitudes and the intricate interplay between travelers’ preferences for MaaS and the uncertainties inherent in daily travel.

Black-Box Optimization (BBO) has found successful applications in many fields of science and engineering. Recently, there has been a growing interest in meta-learning particular components of BBO algorithms to speed up optimization and get rid of tedious hand-crafted heuristics. As an extension, learning the entire algorithm from data requires the least labor from experts and can provide the most flexibility. In this paper, we propose RIBBO, a method to reinforce-learn a BBO algorithm from offline data in an end-to-end fashion. RIBBO employs expressive sequence models to learn the optimization histories produced by multiple behavior algorithms and tasks, leveraging the in-context learning ability of large models to extract task information and make decisions accordingly. Central to our method is to augment the optimization histories with regret-to-go tokens, which are designed to represent the performance of an algorithm based on cumulative regret of the histories. The integration of regret-to-go tokens enables RIBBO to automatically generate sequences of query points that satisfy the user-desired regret, which is verified by its universally good empirical performance on diverse problems, including BBOB functions, hyper-parameter optimization and robot control problems.

Localization and mapping are critical tasks for various applications such as autonomous vehicles and robotics. The challenges posed by outdoor environments present particular complexities due to their unbounded characteristics. In this work, we present MM-Gaussian, a LiDAR-camera multi-modal fusion system for localization and mapping in unbounded scenes. Our approach is inspired by the recently developed 3D Gaussians, which demonstrate remarkable capabilities in achieving high rendering quality and fast rendering speed. Specifically, our system fully utilizes the geometric structure information provided by solid-state LiDAR to address the problem of inaccurate depth encountered when relying solely on visual solutions in unbounded, outdoor scenarios. Additionally, we utilize 3D Gaussian point clouds, with the assistance of pixel-level gradient descent, to fully exploit the color information in photos, thereby achieving realistic rendering effects. To further bolster the robustness of our system, we designed a relocalization module, which assists in returning to the correct trajectory in the event of a localization failure. Experiments conducted in multiple scenarios demonstrate the effectiveness of our method., Comment: 7 pages, 5 figures

In the 21st century, China is experiencing rapid economic growth alongside increasing car ownership and its attendant challenges. However, little is known about the determinants of expenditure on car acquisition and household car ownership in China, especially their dynamics as this motorization process has taken place. In this paper, we present household car ownership and expenditure analyses using a unique panel dataset, the China Household Finance Survey (CHFS). We identify three clusters of households with distinct income and expenditure characteristics consistently existing from 2011 to 2017. We also demonstrate distinctive temporal patterns in expenditure in the periods prior to, at-time-of, and following car acquisition for car-purchasing households. We report that most, but not all, determinants we studied have been stable over time. Finally, contrary to contemporary findings in Western countries, being a younger-aged household is positively associated with car ownership, while living in rural areas has a negative association. © 2023 Elsevier Ltd

Mobility-as-a-Service (MaaS) is a promising solution for sustainable mobility and has therefore received considerable attention from researchers. People have complicated travel diaries. Larger households, especially with children, with a limited number of cars, are essentially prone to delays or travel cancellations. MaaS options fill this gap for such families by offering mobility options accessible at low ownership costs. This paper investigates the impact of non-mandatory trips (mandatory trips include work and school) on MaaS subscription preferences. Different likelihoods of occurrence and the corresponding travel pattern shifts are examined using a stated preference survey of 997 individuals in Hong Kong. A two-stage model is developed: we first calibrate individual risk attitudes using the Constant Relative Risk Aversion (CRRA) utility function. Then we develop several mixed logit models that incorporate risks associated with activities and the response of travelers to the scale of such uncertainty on mode shift decisions. The results reveal that respondents tend to be more multimodal after subscribing to a MaaS bundle, and the use of a taxi is greatly encouraged after subscription. In terms of users’ risk preference, we find that more risk-averse individuals tend to be reluctant to subscribe to MaaS bundles when the level of uncertainty increases. Moreover, our findings demonstrate that travelers are more willing to pay for flexible travel options and have a higher value of time when facing uncertainties (i.e., in the presence of occasional activities). These findings can provide valuable insights for both policymakers and MaaS operators, reflecting users’ perceptions regarding the features of MaaS. © 2023 Hong Kong Society for Transportation Studies

Decision Transformer (DT), which employs expressive sequence modeling techniques to perform action generation, has emerged as a promising approach to offline policy optimization. However, DT generates actions conditioned on a desired future return, which is known to bear some weaknesses such as the susceptibility to environmental stochasticity. To overcome DT's weaknesses, we propose to empower DT with dynamic programming. Our method comprises three steps. First, we employ in-sample value iteration to obtain approximated value functions, which involves dynamic programming over the MDP structure. Second, we evaluate action quality in context with estimated advantages. We introduce two types of advantage estimators, IAE and GAE, which are suitable for different tasks. Third, we train an Advantage-Conditioned Transformer (ACT) to generate actions conditioned on the estimated advantages. Finally, during testing, ACT generates actions conditioned on a desired advantage. Our evaluation results validate that, by leveraging the power of dynamic programming, ACT demonstrates effective trajectory stitching and robust action generation in spite of the environmental stochasticity, outperforming baseline methods across various benchmarks. Additionally, we conduct an in-depth analysis of ACT's various design choices through ablation studies. Our code is available at https://github.com/LAMDA-RL/ACT., Comment: Accepted by AAAI 2024

China is motorizing rapidly as economic growth takes place alongside increasing car ownership. However, little is known about the determinants of household car ownership, and especially about how they are changing as this motorization process takes place. A major reason for this knowledge gap is the very limited available data resources compared to Western societies. In this paper, we present analyses of household car ownership and expenditure on automobiles using a unique panel dataset, the China Household Finance Survey (CHFS). Among our findings, we identify four clusters of households with distinct income and expenditure characteristics consistently existing from 2011 to 2017. We also demonstrate distinctive temporal patterns in expenditure in the periods prior to, at-time-of, and following car acquisition for car-purchasing households. Finally, factors underlying car ownership and expenditure are found to be generally consistent across waves, however, a younger household is revealed to be a positive predictor for car ownership which is contrary to findings in Western countries.

In this study, we investigated whether there is a repetition benefit in mental rotation that is independent of stimulus repetition (i.e., due to increased efficiency in postencoding processing). Three experiments were conducted, in which different conditions of stimulus repetition (different letters on consecutive trials in Experiment 1, letters of different orientations on consecutive trials in Experiment 2, and priming of rotation direction in Experiment 3) were used, and the extent of repetition of rotation direction between two consecutive trials was manipulated. The results of all three experiments showed clear evidence of a repetition benefit without repeating the stimulus, suggesting that this effect is independent of stimulus repetition and lending support to the notion of increased efficiency in mental rotation as a result of repeated rotation direction per se.