HPA* Enhancements

Jansen, Michel; Buro, Michael

doi:10.1609/aiide.v3i1.18791

Cited by 6 publications

(5 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Existing path planning algorithms are capable of effectively completing planning tasks and ensuring the validity of their paths. However, as the environment and map scale up, algorithms like A* and its variants (Hart et al, 1968b;Korf et al, 2001;Harabor and Grastien, 2011;Jansen and Buro, 2007) encounter an exponential increase in computational and memory demands. This occurs because the pathfinding process can become sub-optimal (see Figure 1), where the algorithm might spend unnecessary effort exploring less relevant areas, leading to exponential increases in time complexity as the map size enlarges.…”

Section: Introductionmentioning

confidence: 99%

LLM-A*: Large Language Model Enhanced Incremental Heuristic Search on Path Planning

Meng,

Wang,

Yang

et al. 2024

Preprint

View full text Add to dashboard Cite

Path planning is a fundamental scientific problem in robotics and autonomous navigation, requiring the derivation of efficient routes from starting to destination points while avoiding obstacles. Traditional algorithms like A* and its variants are capable of ensuring path validity but suffer from significant computational and memory inefficiencies as the state space grows. Conversely, large language models (LLMs) excel in broader environmental analysis through contextual understanding, providing global insights into environments. However, they fall short in detailed spatial and temporal reasoning, often leading to invalid or inefficient routes. In this work, we propose LLM-A*, an new LLM based route planning method that synergistically combines the precise pathfinding capabilities of A* with the global reasoning capability of LLMs. This hybrid approach aims to enhance pathfinding efficiency in terms of time and space complexity while maintaining the integrity of path validity, especially in large-scale scenarios. By integrating the strengths of both methodologies, LLM-A* addresses the computational and memory limitations of conventional algorithms without compromising on the validity required for effective pathfinding.

show abstract

Section: Introductionmentioning

confidence: 99%

LLM-A*: Large Language Model Enhanced Incremental Heuristic Search on Path Planning

Meng,

Wang,

Yang

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…Subsequently, the initial and destination points were connected to the graph to create the final path. The method has undergone several improvements [29,30]. HPA* is designed to perform path planning on a nonuniform cost grid and is primarily employed in the fields of robotics and gaming.…”

Section: Introductionmentioning

confidence: 99%

An Effective Method for Computing the Least-Cost Path Using a Multi-Resolution Raster Cost Surface Model

Tang

Dou

2023

IJGI

View full text Add to dashboard Cite

Calculating the least-cost path (LCP) is a fundamental operation in raster-based geographic information systems (GIS). The LCP is applied to raster cost surfaces, in which it determines the most cost-effective path. Increasing the raster resolution results in a longer computation time to obtain LCP. This paper proposes a method for calculating the LCP using a multi-resolution raster cost surface model to enhance computational performance for large-scale grids. The original raster cost surface is progressively downsampled to generate grids of decreasing resolutions. Subsequently, the path is determined on the low-resolution raster. By performing operations such as filtering directional points and mapping path points, the final path on the high-resolution raster can be obtained. The method enables a parallel computation of paths. Therefore, it significantly improves the efficiency for synthetic raster cost surfaces with continuous or discrete characteristics, as well as for raster cost surfaces generated from real terrain datasets, while also providing an end-to-end path output. The experiments show that 80% of the results are very close to the original LCP, and the accuracy of the remaining paths falls within an acceptable range. At the same time, our method greatly improves the efficiency of path solving on a large-scale raster, fulfilling practical application requirements.

show abstract

“…FIGURE 2.17: The HPA* graph inter-cluster and intra-cluster links [Anguelov, 2012] Intra edge searches proves to be costly in hierarchical algorithms, to solve this problem, search for intra edges are postponed with the anticipation that intra edges may not be required in the short run, or the set of cluster data may change before the need of intra edges, avoiding expensive operation which may lead to a costly search in the long run [Jansen and Buro, 2007].…”

Section: Merits Of Using Hierarchical Pathfindingmentioning

confidence: 99%

“…with the pair of start and goal nodes inserted, a refined abstract path is the created. But the use of several A * searches results in more computational costs [Jansen and Buro, 2007].…”

Section: Merits Of Using Hierarchical Pathfindingmentioning

confidence: 99%

“…The computational cost of inserting both start and goal nodes into the abstract graph cab be deductible by using Dijkstra's algorithm rather than A* to compute the shortest path to all bordering abstract nodes as outlined in [Jansen and Buro, 2007]. Therefore [Jansen and Buro, 2007] proposes the use of the Dijkstra's algorithm to find shortest path to all abstract nodes in the quest of finding the nearest reachable abstract node which offers a comparatively cheaper computational burden. Even though this method offers a less cost, the search for shortest paths only works within clusters of many abstract nodes which has complex or long paths between them.…”

Section: Merits Of Using Hierarchical Pathfindingmentioning

confidence: 99%

See 1 more Smart Citation

Toward human-like pathfinding with hierarchical approaches and the GPS of the brain theory

Rahmani

View full text Add to dashboard Cite

Pathfinding for autonomous agents and robots has been traditionally driven by finding optimal paths. Where typically optimality means finding the shortest path between two points in a given environment. However, optimality may not always be strictly necessary. For example, in the case of video games, often computing the paths for non-player characters (NPC) must be done under strict time constraints to guarantee real time simulation. In those cases, performance is more important than finding the shortest path, specially because often a sub-optimal path can be just as convincing from the point of view of the player. When simulating virtual humanoids, pathfinding has also been used with the same goal: finding the shortest path. However, humans very rarely follow precise shortest paths, and thus there are other aspects of human decision making and path planning strategies that should be incorporated in current simulation models. In this thesis we first focus on improving performance optimallity to handle as many virtual agents as possible, and then introduce neuroscience research to propose pathfinding algorithms that attempt to mimic humans in a more realistic manner.In the case of simulating NPCs for video games, one of the main challenges is to compute paths as efficiently as possible for groups of agents. As both the size of the environments and the number of autonomous agents increase, it becomes harder to obtain results in real time under the constraints of memory and computing resources. For this purpose we explored hierarchical approaches for two reasons: (1) they have shown important performance improvements for regular grids and other abstract problems, and (2) humans tend to plan trajectories also following an topbottom abstraction, focusing first on high level location and then refining the path as they move between those high level locations. Therefore, we believe that hierarchical approaches combine the best of our two goals: improving performance for multi-agent pathfinding and achieving more human-like pathfinding. Hierarchical approaches, such as HNA* (Hierarchical A* for Navigation Meshes) can compute paths more efficiently, although only for certain configurations of the hierarchy. For other configurations, the method suffers from a bottleneck in the step that connects the Start and Goal positions with the hierarchy. This bottleneck can drop performance drastically.In this thesis we present different approaches to solve the HNA* bottleneck and thus obtain a performance boost for all hierarchical configurations. The first method relies on further memory storage, and the second one uses parallelism on the GPU. Our comparative evaluation shows that both approaches offer speed-ups as high as 9x faster than A*, and show no limitations based on hierarchical configuration. Then we further exploit the potential of CUDA parallelism, to extend our implementation to HNA* for multi-agent path finding. Our method can now compute paths for over 500K agents simultaneously in real-time, with speed-ups above 15x faster than a parallel multi-agent implementation using A*. We then focus on studying neurosience research to learn about the way that humans build mental maps, in order to propose novel algorithms that take those finding into account when simulating virtual humans. We propose a novel algorithm for path finding that is inspired by neuroscience research on how the brain learns and builds cognitive maps. Our method represents the space as a hexagonal grid, based on the GPS of the brain theory, and fires memory cells as counters. Our path finder then combines a method for exploring unknown environments while building such a cognitive map, with an A* search using a modified heuristic that takes into account the GPS of the brain cognitive map. El problema de Pathfinding para agentes autónomos o robots, ha consistido tradicionalmente en encontrar un camino óptimo, donde por óptimo se entiende el camino de distancia mínima entre dos posiciones de un entorno. Sin embargo, en ocasiones puede que no sea estrictamente necesario encontrar una solución óptima. Para ofrecer la simulación de multitudes de agentes autónomos moviéndose en tiempo real, es necesario calcular sus trayectorias bajo condiciones estrictas de tiempo de computación, pero no es fundamental que las soluciones sean las de distancia mínima ya que, con frecuencia, el observador no apreciará la diferencia entre un camino óptimo y un sub-óptimo. Por tanto, suele ser suficiente con que la solución encontrada sea visualmente creíble para el observado. Cuando se simulan humanoides virtuales en aplicaciones de realidad virtual que requieren avatares que simulen el comportamiento de los humanos, se tiende a emplear los mismos algoritmos que en video juegos, con el objetivo de encontrar caminos de distancia mínima. Pero si realmente queremos que los avatares imiten el comportamiento humano, tenemos que tener en cuenta que, en el mundo real, los humanos rara vez seguimos precisamente el camino más corto, y por tanto se deben considerar otros aspectos que influyen en la toma de decisiones de los humanos y la selección de rutas en el mundo real. En esta tesis nos centraremos primero en mejorar el rendimiento de la búsqueda de caminos para poder simular grandes números de humanoides virtuales autónomos, y a continuación introduciremos conceptos de investigación con mamíferos en neurociencia, para proponer soluciones al problema de pathfinding que intenten imitar con mayor realismo, el modo en el que los humanos navegan el entorno que les rodea. A medida que aumentan tanto el tamaño de los entornos virtuales como el número de agentes autónomos, resulta más difícil obtener soluciones en tiempo real, debido a las limitaciones de memoria y recursos informáticos. Para resolver este problema, en esta tesis exploramos enfoques jerárquicos porque consideramos que combinan dos objetivos fundamentales: mejorar el rendimiento en la búsqueda de caminos para multitudes de agentes y lograr una búsqueda de caminos similar a la de los humanos. El primer método presentado en esta tesis se basa en mejorar el rendimiento del algoritmo HNA* (Hierarchical A* for Navigation Meshes) incrementando almacenamiento de datos en memoria, y el segundo utiliza el paralelismo para mejorar drásticamente el rendimiento. La evaluación cuantitativa realizada en esta tesis, muestra que ambos enfoques ofrecen aceleraciones que pueden llegar a ser hasta 9 veces más rápidas que el algoritmo A* y no presentan limitaciones debidas a la configuración jerárquica. A continuación, aprovechamos aún más el potencial del paralelismo ofrecido por CUDA para extender nuestra implementación de HNA* a sistemas multi-agentes. Nuestro método permite calcular caminos simultáneamente y en tiempo real para más de 500.000 agentes, con una aceleración superior a 15 veces la obtenida por una implementación paralela del algoritmo A*. Por último, en esta tesis nos hemos centrado en estudiar los últimos avances realizados en el ámbito de la neurociencia, para comprender la manera en la que los humanos construyen mapas mentales y poder así proponer nuevos algoritmos que imiten de forma más real el modo en el que navegamos los humanos. Nuestro método representa el espacio como una red hexagonal, basada en la distribución de ¿place cells¿ existente en el cerebro, e imita las activaciones neuronales como contadores en dichas celdas. Nuestro buscador de rutas combina un método para explorar entornos desconocidos mientras construye un mapa cognitivo hexagonal, utilizando una búsqueda A* con una nueva heurística adaptada al mapa cognitivo del cerebro y sus contadores.

show abstract

HPA* Enhancements

Cited by 6 publications

References 0 publications

LLM-A*: Large Language Model Enhanced Incremental Heuristic Search on Path Planning

LLM-A*: Large Language Model Enhanced Incremental Heuristic Search on Path Planning

An Effective Method for Computing the Least-Cost Path Using a Multi-Resolution Raster Cost Surface Model

Toward human-like pathfinding with hierarchical approaches and the GPS of the brain theory

Contact Info

Product

Resources

About