Abstract
Combining motion prediction and motion planning offers a promising framework for enhancing interactions between automated vehicles and other traffic participants. However, this introduces challenges in conditioning predictions on navigation goals and ensuring stable, kinematically feasible trajectories. Addressing the former challenge, this paper investigates the extension of attention-based motion prediction models with navigation information. By integrating the ego vehicle's intended route and goal pose into the model architecture, we bridge the gap between multi-agent motion prediction and goal-based motion planning. We propose and evaluate several architectural navigation integration strategies to our model on the nuPlan dataset. Our results demonstrate the potential of prediction-driven motion planning, highlighting how navigation information can enhance both prediction and planning tasks. Our implementation is at: https://github.com/KIT-MRT/future-motion.

AI Summary

• Navigation information: Information about the vehicle's current location, velocity, and trajectory, as well as the locations and trajectories of other vehicles in the scene. [3]
• The paper discusses the importance of incorporating navigation information into motion forecasting models for autonomous driving. [2]

Abstract
Information gathering in large-scale or time-critical scenarios (e.g., environmental monitoring, search and rescue) requires broad coverage within limited time budgets, motivating the use of multi-agent systems. These scenarios are commonly formulated as multi-agent informative path planning (MAIPP), where multiple agents must coordinate to maximize information gain while operating under budget constraints. A central challenge in MAIPP is ensuring effective coordination while the belief over the environment evolves with incoming measurements. Recent learning-based approaches address this by using distributions over future positions as "intent" to support coordination. However, these autoregressive intent predictors are computationally expensive and prone to compounding errors. Inspired by the effectiveness of diffusion models as expressive, long-horizon policies, we propose AID, a fully decentralized MAIPP framework that leverages diffusion models to generate long-term trajectories in a non-autoregressive manner. AID first performs behavior cloning on trajectories produced by existing MAIPP planners and then fine-tunes the policy using reinforcement learning via Diffusion Policy Policy Optimization (DPPO). This two-stage pipeline enables the policy to inherit expert behavior while learning improved coordination through online reward feedback. Experiments demonstrate that AID consistently improves upon the MAIPP planners it is trained from, achieving up to 4x faster execution and 17% increased information gain, while scaling effectively to larger numbers of agents. Our implementation is publicly available at https://github.com/marmotlab/AID.

Abstract
Large Language Models (LLMs) have become foundational tools in artificial intelligence, supporting a wide range of applications beyond traditional natural language processing, including urban systems and tourist recommendations. However, their tendency to hallucinate and their limitations in spatial retrieval and reasoning are well known, pointing to the need for novel solutions. Retrieval-augmented generation (RAG) has recently emerged as a promising way to enhance LLMs with accurate, domain-specific, and timely information. Spatial RAG extends this approach to tasks involving geographic understanding. In this work, we introduce WalkRAG, a spatial RAG-based framework with a conversational interface for recommending walkable urban itineraries. Users can request routes that meet specific spatial constraints and preferences while interactively retrieving information about the path and points of interest (POIs) along the way. Preliminary results show the effectiveness of combining information retrieval, spatial reasoning, and LLMs to support urban discovery.

AI Summary

• The method is evaluated on various benchmarks, including the StepGame benchmark and the MS MARCO dataset. [3]
• The paper introduces a novel approach to spatial reasoning in large language models using retrieval-augmented generation. [2]
• The results show that the proposed approach outperforms state-of-the-art methods in terms of accuracy and efficiency. [1]

Abstract
Integrating demand-responsive mobility services with transit systems is recognized as a practical and effective strategy to mitigate their impact on traffic congestion and the environment. This study develops an efficient hybrid metaheuristic to solve the integrated dial-a-ride problem by utilizing electric vehicles to minimize operational costs and customer travel time. Customer transfer inconvenience is restricted by a maximum intermodal transfer time to synchronize demand-responsive buses' arrival and transit departures. The proposed metaheuristic addresses the challenges of integrating demand-responsive vehicle routing and charging operations with fixed-route transit systems with capacitated charging stations and partial recharge. We benchmarked our algorithm against a state-of-the-art mixed-integer programming solver on instances with 10-50 customers and two transit lines. Our approach achieves solutions that are, on average, 23.8% better in solution quality within around 2 minutes, outperforming those obtained by the solver using an 8-hour computational time limit. We evaluate the impact of various system parameters to bridge the gap between theory and practice. The results suggest that, from the operator's perspective, while the integrated dial-a-ride service reduces vehicle kilometers traveled, the used fleet size may not necessarily be reduced when ensuring high-quality service for passengers. Moreover, operating the integrated systems is more beneficial in areas with dense transit networks, compared with increases in transit frequency. The findings provide valuable insights for developing integrated dial-a-ride services in practice.

Abstract
The "15-minute city" envisions neighborhoods where residents can meet daily needs via a short walk or bike ride. Realizing this vision requires not only physical proximity but also efficient and reliable access to information about nearby places, services, and events. Existing location-based systems, however, focus mainly on city-level tasks and neglect the spatial, temporal, and cognitive factors that shape localized decision-making. We conceptualize this gap as the Local Life Information Accessibility (LLIA) problem and introduce AskNearby, an AI-driven community application that unifies retrieval and recommendation within the 15-minute life circle. AskNearby integrates (i) a three-layer Retrieval-Augmented Generation (RAG) pipeline that synergizes graph-based, semantic-vector, and geographic retrieval with (ii) a cognitive-map model that encodes each user's neighborhood familiarity and preferences. Experiments on real-world community datasets demonstrate that AskNearby significantly outperforms LLM-based and map-based baselines in retrieval accuracy and recommendation quality, achieving robust performance in spatiotemporal grounding and cognitive-aware ranking. Real-world deployments further validate its effectiveness. By addressing the LLIA challenge, AskNearby empowers residents to more effectively discover local resources, plan daily activities, and engage in community life.

AI Summary

• AskNearby is an AI-driven community platform that integrates a three-layer retrieval-augmented generation framework with a cognitive map-based recommendation model. [3]
• The paper introduces the problem of Localized Large-scale Information Access (LLIA), which focuses on enabling residents to efficiently acquire relevant, timely, and context-aware information within their neighborhood. [2]