Algorithms for the computation of alternative routes in road networks power many geographic navigation systems. A good set of alternative routes offers meaningful options to the user of the system and can support applications such as routing that is robust to failures (e.g., road closures, extreme traffic congestion, etc.) and routing with diverse preferences and objective functions. Algorithmic techniques for alternative route computation include the penalty method, via-node type algorithms (which deploy bidirectional search and finding plateaus), and, more recently, electrical-circuit based algorithms. In this work we focus on the practically important family of via-node type algorithms and aim to produce high quality alternative routes for road networks using a novel deep learning-based approach that learns a representation of the underlying road network. We show that this approach can support natural objectives, such as the uniformly bounded stretch, that are difficult and computationally expensive to support through traditional algorithmic techniques. Moreover, we achieve this in a practical system based on the Customizable Route Planning (CRP) hierarchical routing architecture. Our training methodology uses the hierarchical partition of the graph and trains a model to predict which boundary nodes in the partition should be crossed by the alternative routes. We describe our methods in detail and evaluate them against previously studied baselines, showing quality improvements in the road networks of Seattle, Paris, and Bangalore.

We study learning-based approaches to semantic route planning, which concerns producing routes in response to rich queries that specify various criteria and preferences. Semantic routing is already widely found in industry applications, especially navigational services like Google Maps; however, existing implementations only support limited route criteria and narrow query sets as they rely on repurposing classical route optimization algorithms. We argue for a learning-based approach to semantic routing as a more scalable and general alternative. To foster interest in this important application of graph learning, we are releasing a large-scale publicly-licensed benchmark for semantic routing consisting of real-world multi-objective navigation problems---expressed via natural language queries---on the richly annotated road networks of US cities. In addition to being intractable with existing approaches to semantic routing, our benchmark poses a significant scaling challenge for graph learning methods. As a proof-of-concept, we show that---at scale---even a standard transformer network is a powerful semantic routing system and achieves non-trivial performance on our benchmark. In the process, we demonstrate a simple solution to the challenge of scaling up graph learning: an autoregressive approach that decomposes semantic routing into smaller ``next-edge'' prediction problems.

Electric Vehicle routing is often modeled as a Shortest Feasible Path Problem (SFPP), which minimizes total travel time while maintaining a non-zero State of Charge (SoC) along the route. However, the problem assumes perfect information about energy consumption and charging stations, which are difficult to even estimate in practice. Further, drivers might have varying risk tolerances for different trips. To overcome these limitations, we propose two generalizations to the SFPP; they compute the shortest feasible path for any initial SoC and, respectively, for every possible minimum SoC threshold. We present algorithmic solutions for each problem, and provide two constructs: Starting Charge Maps and Buffer Maps, which represent the tradeoffs between robustness of feasible routes and their travel times. The two constructs are useful in many ways, including presenting alternate routes or providing charging prompts to users. We evaluate the performance of our algorithms on realistic input instances.

We study the problem of graph partitioning for evolving road networks. While the road network of the world is mostly stable, small updates happen on a relatively frequent basis, as can been observed with the OpenStreetMap project (http://www.openstreetmap.org). For various reasons, professional applications demand the graph partition to stay roughly the same over time, and that changes are limited to areas where graph updates occur. In this work, we define the problem, present algorithms to satisfy the stability needs, and evaluate our techniques on continental-sized road networks. Besides the stability gains, we show that, when the changes are low and local, running our novel techniques is an order of magnitude faster than running graph partitioning from scratch.

We present a novel exact journey planning approach to computing a reasonable subset of multi-criteria Pareto sets in public transit networks. Our restriction is well defined and independent of the choice of algorithm. In order to compute the restricted Pareto set efficiently, we present Bounded McRAPTOR, a new set of algorithms that extend the well-known McRAPTOR algorithm. The fastest variant employs a novel pruning scheme based on carefully computed bounds. Experiments on large metropolitan networks show that a four-criteria restricted Pareto set can be computed faster by a factor of up to 65, while retaining the important journeys of the full Pareto set. This easily enables interactive applications in practice, making multi-criteria Pareto-optimal journey planning scalable without the need of a preprocessing-based speedup technique.

The last 15 years have seen astonishing progress in the performance of shortest path algorithms for transportation networks. In particular, for road networks, modern algorithms can be up to seven orders of magnitude faster than standard solutions. Since these algorithms enable several new applications, many of them have found their way into navigation services of major technology companies serving hundreds of millions of users every day. This talk highlights key techniques, discusses their impact on the industry, and provides an outlook on upcoming challenges.

We study how to compute routes that avoid traffic in road networks. Imperfections in real-time traffic feeds may yield routes with undesirable detours through parking lots or residential areas. The main challenge we address in this work is that of defining and computing paths that incorporate a volatile secondary cost function. We define the problem, study its complexity, and present algorithms that compute routes without undesirable detours. Experiments on continental-sized road networks demonstrate the feasibility of our approach.

We present a preprocessing-based acceleration technique for computing bi-criteria Pareto-optimal journeys in public transit networks, based on the well-known RAPTOR algorithm [Delling et al 2015]. Our key idea is to first partition a hypergraph into cells, in which vertices correspond to routes (e.g., bus lines) and hyperedges to stops, and to then mark routes sufficient for optimal travel across cells. The query can then be restricted to marked routes and those in the source and target cells. This results in a practical approach, suitable for networks that are too large to be efficiently handled by the basic RAPTOR algorithm.

Given a base weighted planar graph Ginput on n nodes and parameters M, ∊ we present a dynamic distance oracle with 1 + ∊ stretch and worst case update and query costs of ∊–3M4 · poly-log(n). We allow arbitrary edge weight updates as long as the shortest path metric induced by the updated graph has stretch of at most M relative to the shortest path metric of the base graph Ginput. For example, on a planar road network, we can support fast queries and dynamic traffic updates as long as the shortest path from any source to any target (including using arbitrary detours) is between, say, 80 and 3 miles-per-hour. As a warm-up we also prove that graphs of bounded treewidth have exact distance oracles in the dynamic edge model. To the best of our knowledge, this is the first dynamic distance oracle for a non-trivial family of dynamic changes to planar graphs with worst case costs of o(n1/2) both for query and for update operations.

Computing driving directions has motivated many shortest path algorithms based on preprocessing. Given a graph, the preprocessing stage computes a modest amount of auxiliary data, which is then used to speed up online queries. In practice, the best algorithms have storage overhead comparable to the graph size and answer queries very fast, while examining a small fraction of the graph. In this article, we complement the experimental evidence with the first rigorous proofs of efficiency for some of the speedup techniques developed over the past decade or variations thereof. We define highway dimension, which strengthens the notion of doubling dimension. Under the assumption that the highway dimension is low (at most polylogarithmic in the graph size), we show that, for some algorithms or their variants, preprocessing can be implemented in polynomial time, the resulting auxiliary data increases the storage requirements by a polylogarithmic factor, and queries run in polylogarithmic time. This gives a unified explanation for the performance of several seemingly different approaches. Our best bounds are based on a result that may be of independent interest: we show that unique shortest paths induce set systems of low VC-dimension, which makes them combinatorially simple.

All current navigation systems return efficient source-to-destination routes assuming a "one-size-fits-all" set of objectives, without addressing most personal preferences. Although they allow some customization (like "avoid highways" or "avoid tolls"), the choices are very limited and require some sophistication on the part of the user. In this paper we present, implement, and test a framework that generates personalized driving directions by automatically analyzing users' GPS traces. Our approach learns cost functions using coordinate descent, leveraging a state-of-the-art route planning engine for efficiency. In an extensive experimental study, we show that this framework infers user-specific driving preferences, significantly improving the route quality. Our approach can handle continental-sized inputs (with tens of millions of vertices and arcs) and is efficient enough to be run on an autonomous device (such as a car navigation system) preserving user privacy.

We present a unified framework for dealing with exact point-of-interest (POI) queries in dynamic continental road networks within interactive applications. We show that partition-based algorithms developed for point-to-point shortest path computations can be naturally extended to handle augmented queries such as finding the closest restaurant or the best post office to stop on the way home, always ranking POIs according to a user-defined cost function. Our solution allows different trade-offs between indexing effort (time and space) and query time. Our most flexible variant allows the road network to change frequently (to account for traffic information or personalized cost functions) and the set of POIs to be specified at query time. Even in this fully dynamic scenario, our solution is fast enough for interactive applications on continental road networks.

Propagation of contagion through networks is a fundamental process. It is used to model the spread of information, influence, or a viral infection. Diffusion patterns can be specified by a probabilistic model, such as Independent Cascade (IC), or captured by a set of representative traces.

We present a unified framework for exact point-of-interest (POI) queries in dynamic road networks. We show how to extend partition-based shortest-path algorithms to answer queries such as finding the closest restaurant or the best post office to stop on the way home, ranking POIs according to a user-defined cost function. We provide various trade-offs between indexing effort and query time. Our most flexible variant allows the road network to change frequently (to account for traffic or personalized cost functions) and the set of POIs to be specified at query time, and is still fast enough for interactive applications on continental networks.

Computing driving directions in road networks is a fundamental problem. Although it can be solved in essentially linear time by Dijkstra's algorithm, this is not fast enough to enable interactive queries on large-scale inputs. Instead, modern algorithms typically work in two stages: first an offline preprocessing routine computes some auxiliary data, which is then used to answer exact queries in real time. The past decade has seen a surprisingly diverse set of techniques that follow this approach, mostly relying on the fact that road networks tend to have a strong hierarchy. These methods work very well when minimizing driving times, but are much less efficient with other cost functions. We present a practical algorithm that has no such drawbacks, and can compute shortest paths on continental road networks with arbitrary metrics (cost functions). Our customizable route planning approach works in three stages. The first, metric-independent preprocessing, uses graph partitioning to define the topology of a multilevel overlay graph, which is the same regardless of the cost function. The second stage, customization, uses the metric to compute the actual costs of the overlay arcs. Finally, the query stage uses the output of the first two stages to compute shortest paths in real time (milliseconds). The first stage uses a recent partitioning algorithm based on the notion of natural cuts, which are sparse regions separating much denser areas. It may take a few minutes (or even hours), but only needs to be run (or updated) when new road segments are built. Metric changes (which are much more frequent) require running only customization, which takes a second or less even on continental road networks. Since it does not rely on strong hierarchies, CRP is robust to metric changes. Unlike most other methods, it can also handle turn costs (and restrictions) quite naturally, with little effect on performance and space usage. It is thus ideal for a real-world routing engine, and is indeed in use by Bing Maps. This extended abstract includes results first published at SEA 2011 and SEA 2013.

We study the problem of computing all Pareto-optimal journeys in a dynamic public transit network for two criteria: arrival time and number of transfers. Existing algorithms consider this as a graph problem, and solve it using variants of Dijkstra's algorithm. Unfortunately, this leads to either high query times or suboptimal solutions. We take a different approach. We introduce RAPTOR, our novel round-based public transit router. Unlike previous algorithms, it is not Dijkstra-based, looks at each route (such as a bus line) in the network at most once per round, and can be made even faster with simple pruning rules and parallelization using multiple cores. Because it does not rely on preprocessing, RAPTOR works in fully dynamic scenarios. Moreover, it can be easily extended to handle flexible departure times or arbitrary additional criteria, such as fare zones. When run on London's complex public transportation network, RAPTOR computes all Pareto-optimal journeys between two random locations an order of magnitude faster than previous approaches, which easily enables interactive applications. This is an extended abstract of the paper published at ALENEX 2012.

We present a novel exact algorithm for the minimum graph bisection problem, whose goal is to partition a graph into two equally-sized cells while minimizing the number of edges between them. Our algorithm is based on the branch-and-bound framework and, unlike most previous approaches, it is fully combinatorial. We present stronger lower bounds, improved branching rules, and a new decomposition technique that contracts entire regions of the graph without losing optimality guarantees. In practice, our algorithm works particularly well on instances with relatively small minimum bisections, solving large real-world graphs (with tens of thousands to millions of vertices) to optimality.

We study the problem of route planning on mobile devices. There are two current approaches to this problem. One option is to have all the routing data on the device, which can then compute routes by itself. This makes it hard to incorporate traffic updates, leading to suboptimal routes. An alternative approach outsources the route computation to a server, which then sends only the route to the device. The downside is that a user is lost when deviating from the proposed route in an area with limited connectivity. In this work, we present an approach that combines the best of both worlds. The server performs the route computation but, instead of sending only the route to the user, it sends a corridor that is robust against deviations. We define these corridors properly and show that their size can be theoretically bounded in road networks. We evaluate their quality experimentally in terms of size and robustness on a continental road network. Finally, we introduce several algorithms to compute corridors efficiently. Our experimental analysis shows that our corridors are small but very robust against deviations, and can be computed quickly on a standard server.

We study the problem of computing all Pareto-optimal journeys in a dynamic public transit network for two criteria: arrival time and number of transfers.Existing algorithms consider this as a graph problem, and solve it using variants of Dijkstra's algorithm.Unfortunately, this leads to either high query times or suboptimal solutions.We take a different approach.We introduce RAPTOR, our novel round-based public transit router.Unlike previous algorithms, it is not Dijkstrabased, looks at each route (such as a bus line) in the network at most once per round, and can be made even faster with simple pruning rules and parallelization using multiple cores.Because it does not rely on preprocessing, RAPTOR works in fully dynamic scenarios.Moreover, it can be easily extended to handle flexible departure times or arbitrary additional criteria, such as fare zones.When run on London's complex public transportation network, RAPTOR computes all Paretooptimal journeys between two random locations an order of magnitude faster than previous approaches, which easily enables interactive applications.

This paper introduces HLDB, the first practical system that can answer exact spatial queries on continental road networks entirely within a database. HLDB is based on hub labels (HL), the fastest point-to-point algorithm for road networks, and its queries are implemented (quite naturally) in standard SQL. Within the database, HLDB answers exact distance queries and retrieves full shortest-path descriptions in real time, even on networks with tens of millions of vertices. The basic algorithm can be extended in a natural way (still in SQL) to answer much more sophisticated queries, such as finding the ten closest fast-food restaurants. We also introduce efficient new HL-based algorithms for even harder problems, such as best via point, ride sharing, and point of interest prediction. The HLDB framework makes it easy to implement these algorithms in SQL, enabling interactive applications on continental road networks.

Route planning in large-scale time-dependent road networks is an important practical application of the shortest-path problem that greatly benefits from speedup techniques. In this paper, we extend a two-level hierarchical approach for point-to-point shortest-path computations to the time-dependent case. This method, also known as core routing in the literature for static graphs, consists of the selection of a small subnetwork where most of the computations can be carried out, thus reducing the search space. We combine this approach with bidirectional goal-directed search to obtain an algorithm capable of finding shortest paths in a matter of milliseconds on continental-sized networks. Moreover, we tackle the dynamic scenario where the piecewise linear functions that we use to model time-dependent arc costs are not fixed but can have their coefficients updated requiring only a small computational effort.

Abstract The computation of point‐to‐point shortest paths on time‐dependent road networks has a large practical interest, but very few works propose efficient algorithms for this problem. We propose a novel approach, which tackles one of the main complications of route planning in time‐dependent graphs, which is the difficulty of using bidirectional search: because the exact arrival time at the destination is unknown, we start a backward search from the destination node using lower bounds on arc costs to restrict the set of nodes that have to be explored by the forward search. Our algorithm is based onA* with landmarks (ALT); extensive computational results show that it is very effective in practice if we are willing to accept a small approximation factor, resulting in a speed‐up of more than one order of magnitude with respect to Dijkstra's algorithm while finding only slightly suboptimal solutions. The main idea presented here can also be generalized to other types of search algorithms. © 2011 Wiley Periodicals, Inc. NETWORKS, 2012

We study the problem of computing batched shortest paths in road networks efficiently. Our focus is on computing paths from a single source to multiple targets (one-to-many queries). We perform a comprehensive experimental comparison of several approaches, including new ones. We conclude that a new extension of PHAST (a recent one-to-all algorithm), called RPHAST, has the best performance in most cases, often by orders of magnitude. When used to compute distance tables (many-to-many queries), RPHAST often outperforms all previous approaches.

We introduce Dryad Opt, a library that enables massively parallel and distributed execution of optimization algorithms for solving hard problems. Dryad Opt performs an exhaustive search of the solution space using branch-and-bound, by recursively splitting the original problem into many simpler sub problems. It uses both parallelism (at the core level) and distributed execution (at the machine level). Dryad Opt provides a simple yet powerful interface to its users, who only need to implement sequential code to process individual sub problems (either by solving them in full or generating new sub problems). The parallelism and distribution are handled automatically by Dryad Opt, and are invisible to the user. The distinctive feature of our system is that it is implemented on top of Dryad LINQ, a distributed data-parallel execution engine similar to Hadoop and Map-Reduce. Despite the fact that these engines offer a constrained application model, with restricted communication patterns, our experiments show that careful design choices allow Dryad Opt to scale linearly with the number of machines, with very little overhead.

We present a novel algorithm to solve the nonnegative single-source shortest path problem on road networks and other graphs with low highway dimension. After a quick preprocessing phase, we can compute all distances from a given source in the graph with essentially a linear sweep over all vertices. Because this sweep is independent of the source, we are able to reorder vertices in advance to exploit locality. Moreover, our algorithm takes advantage of features of modern CPU architectures, such as SSE and multi-core. Compared to Dijkstra's algorithm, our method needs fewer operations, has better locality, and is better able to exploit parallelism at multi-core and instruction levels. We gain additional speedup when implementing our algorithm on a GPU, where our algorithm is up to three orders of magnitude faster than Dijkstra's algorithm on a high-end CPU. This makes applications based on all-pairs shortest-paths practical for continental-sized road networks. Several algorithms, such as computing the graph diameter, exact arc flags, or centrality measures (exact reaches or betweenness), can be greatly accelerated by our method.

We present a novel approach to graph partitioning based on the notion of cuts. Our algorithm, called PUNCH, has two phases. The first phase performs a series of minimum-cut computations to identify and contract dense regions of the graph. This reduces the graph size, but preserves its general structure. The second phase uses a combination of greedy and local search heuristics to assemble the final partition. The algorithm performs especially well on road networks, which have an abundance of natural cuts (such as bridges, mountain passes, and ferries). In a few minutes, it obtains the best known partitions for continental-sized networks, significantly improving on previous results.

Abstract In recent years, many speed up techniques for DIJKSTRA Algorithm have been developed. Unfortunately, research mainly focused on road networks although fast algorithms are also needed for other applications like timetable information systems. Even worse, the adaption of recently developed techniques to graphs deriving from timetable information problems is often more complicated than expected. In this work, we check whether results from road networks are transferable to timetable information systems. To this end, we present an extensive experimental study of the most prominent speed up techniques on inputs deriving from different applications. It turns out that recently developed techniques are much slower on graphs derived from timetable information problems than on road networks. In addition, we gain interesting insights into the behavior of speed up techniques in general. © 2010 Wiley Periodicals, Inc. NETWORKS, 2011

Exploiting parallelism in route planning algorithms is a challenging algorithmic problem with obvious applications in mobile navigation and timetable information systems. In this work, we present a novel algorithm for the so-called one-to-all profile-search problem in public transportation networks. It answers the question for all fastest connections between a given station S and any other station at any time of the day in a single query. This algorithm allows for a very natural parallelization, yielding excellent speed-ups on standard multi-core servers. Our approach exploits the facts that first, time-dependent travel-time functions in such networks can be represented as a special class of piecewise linear functions, and that second, only few connections from S are useful to travel far away. Introducing the connection-setting property, we are able to extend DIJKSTRA's algorithm in a sound manner. Furthermore, we also accelerate station-tostation queries by preprocessing important connections within the public transportation network. As a result, we are able to compute all relevant connections between two random stations in a complete public transportation network of a big city (Los Angeles) on a standard multi-core server in less than 55 ms on average.

Contraction hierarchies are a simple hierarchical routing technique that has proved extremely efficient for static road networks. We explain how to generalize them to networks with time-dependent edge weights. This is the first hierarchical speedup technique for time-dependent routing that allows bidirectional query algorithms. For large realistic networks with considerable time-dependence (Germany, weekdays) our method outperforms previous techniques with respect to query time using comparable or lower preprocessing time.

During the last years, impressive speed-up techniques for Dijkstra's algorithm have been developed. Unfortunately, the most advanced techniques use bidirectional search which makes it hard to use them in scenarios where a backward search is prohibited. Even worse, such scenarios are widely spread, e.g., timetable-information systems or time-dependent networks. In this work, we present a unidirectional speed-up technique which competes with bidirectional approaches. Moreover, we show how to exploit the advantage of unidirectional routing for fast exact queries in timetable information systems and for fast approximative queries in time-dependent scenarios. By running experiments on several inputs other than road networks, we show that our approach is very robust to the input.

Modularity is a recently introduced quality measure for graph clusterings. It has immediately received considerable attention in several disciplines, particularly in the complex systems literature, although its properties are not well understood. We study the problem of finding clusterings with maximum modularity, thus providing theoretical foundations for past and present work based on this measure. More precisely, we prove the conjectured hardness of maximizing modularity both in the general case and with the restriction to cuts and give an Integer Linear Programming formulation. This is complemented by first insights into the behavior and performance of the commonly applied greedy agglomerative approach.