Carlos Hernández

Avoiding and escaping depressions in real-time heuristic search

Heuristics used for solving hard real-time search problems have regions with depressions. Such regions are bounded areas of the search space in which the heuristic function is inaccurate compared to the actual cost to reach a solution. Early real-time search algorithms, like LRTA*, easily become trapped in those regions since the heuristic values of their states may need to be updated multiple times, which results in costly solutions. State-of-the-art real-time search algorithms, like LSS-LRTA* or LRTA*(k), improve LRTA*s mechanism to update the heuristic, resulting in improved performance. Those algorithms, however, do not guide search towards avoiding depressed regions. This paper presents depression avoidance, a simple real-time search principle to guide search towards avoiding states that have been marked as part of a heuristic depression. We propose two ways in which depression avoidance can be implemented: mark-and-avoid and move-to-border. We implement these strategies on top of LSS-LRTA* and RTAA*, producing 4 new real-time heuristic search algorithms: aLSS-LRTA*, daLSS-LRTA*, aRTAA*, and daRTAA*. When the objective is to find a single solution by running the real-time search algorithm once, we show that daLSS-LRTA* and daRTAA* outperform their predecessors sometimes by one order of magnitude. Of the four new algorithms, daRTAA* produces the best solutions given a fixed deadline on the average time allowed per planning episode. We prove all our algorithms have good theoretical properties: in finite search spaces, they find a solution if one exists, and converge to an optimal after a number of trials.

Real-time heuristic search with depression avoidance

Heuristics used for solving hard real-time search problems have regions with depressions. Such regions are bounded areas of the search space in which the heuristic function is exceedingly low compared to the actual cost to reach a solution. Real-time search algorithms easily become trapped in those regions since the heuristic values of states in them may need to be updated multiple times, which results in costly solutions. State-of-theart real-time search algorithms like LSS-LRTA*, LRTA*(k), etc., improve LRTA*s mechanism to update the heuristic, resulting in improved performance. Those algorithms, however, do not guide search towards avoiding or escaping depressed regions. This paper presents depression avoidance, a simple real-time search principle to guide search towards avoiding states that have been marked as part of a heuristic depression. We apply the principle to LSS-LRTA* producing aLSS-LRTA*, a new real-time search algorithm whose search is guided towards exiting regions with heuristic depressions. We show our algorithm outperforms LSS-LRTA* in standard real-time benchmarks. In addition we prove aLSS-LRTA* has most of the good theoretical properties of LSS-LRTA*.

Towards Annotation-Based Query and Document Expansion for Image Retrieval

In this paper we report results of experiments conducted with strategies for improving text-based image retrieval. The adopted strategies were evaluated in the photographic retrieval task at ImageCLEF2007. We propose a Web-based method for expanding textual queries with related terms. This technique was the top-ranked query expansion method among those proposed by other ImageCLEF2007 participants. We also consider two methods for combining visual and textual information in the retrieval process: late-fusion and intermedia-feedback. The best results were obtained by combining intermedia-feedback and our expansion technique. The main contribution of this paper, however, is the proposal of annotation-based expansion; a novel approach that consists of using labels assigned to images (with image annotation methods) for expanding textual queries and documents. We introduce this idea and report results of initial experiments towards enhancing text-based image retrieval via image annotation. Preliminary results show that this expansion strategy could be useful for image retrieval in the near future.

Fast Algorithm for Catching a Prey Quickly in Known and Partially Known Game Maps

In moving target search, the objective is to guide a hunter agent to catch a moving prey. Even though in game applications maps are always available at developing time, current approaches to moving target search do not exploit preprocessing to improve search performance. In this paper, we propose MtsCopa, an algorithm that exploits precomputed information in the form of compressed path databases (CPDs), and that is able to guide a hunter agent in both known and partially known terrain. CPDs have previously been used in standard, fixed-target pathfinding but had not been used in the context of moving target search. We evaluated MtsCopa over standard game maps. Our speed results are orders of magnitude better than current state of the art. The time per individual move is improved, which is important in real-time search scenarios, where the time available to make a move is limited. Compared to state of the art, the number of hunter moves is often better and otherwise comparable, since CPDs provide optimal moves along shortest paths. Compared to previous successful methods, such as I-ARA*, our method is simple to understand and implement. In addition, we prove MtsCopa always guides the agent to catch the prey when possible.

Reusing cost-minimal paths for goal-directed navigation in partially known terrains

Situated agents frequently need to solve search problems in partially known terrains in which the costs of the arcs of the search graphs can increase (but not decrease) when the agents observe new information. An example of such search problems is goal-directed navigation with the freespace assumption in partially known terrains, where agents repeatedly follow cost-minimal paths from their current locations to given goal locations. Incremental heuristic search is an approach for solving the resulting sequences of similar search problems potentially faster than with classical heuristic search, by reusing information from previous searches to speed up its current search. There are two classes of incremental heuristic search algorithms, namely those that make the

Reconnection with the ideal tree: a new approach to real-time search

h

Time-bounded best-first search for reversible and non-reversible search graphs

h-values of the current search more informed (such as Adaptive A*) and those that reuse parts of the A* search trees of previous searches during the current search (such as D* Lite). In this article, we introduce Path-Adaptive A* and its generalization Tree-Adaptive A*. Both incremental heuristic search algorithms terminate their searches before they expand the goal state, namely when they expand a state that is on a provably cost-minimal path to the goal. Path-Adaptive A* stores a single cost-minimal path to the goal state (the reusable path), while Tree-Adaptive A* stores a set of cost-minimal paths to the goal state (the reusable tree), and is thus potentially more efficient than Path-Adaptive A* since it uses information from all previous searches and not just the last one. Tree-Adaptive A* is the first incremental heuristic search algorithm that combines the principles of both classes of incremental heuristic search algorithms. We demonstrate experimentally that both Path-Adaptive A* and Tree-Adaptive A* can be faster than Adaptive A* and D* Lite, two state-of-the-art incremental heuristic search algorithms for goal-directed navigation with the freespace assumption.

Escaping Heuristic Hollows in Real-Time Search without Learning

Multiplying the heuristic function by a weight greater than one is a well-known technique in heuristic search. When this technique is applied to A* with an admissible heuristic it yields substantial runtime savings, at the expense of sacrificing solution optimality. Its applicability to real-time heuristic search, a search approach that builds upon heuristic search, however, has only been explored by a few studies. In this article we present two new approaches to using weights in real-time heuristic search, applicable to a wide range of algorithms. The first one, weighted lookahead, is a variant of an existing approach by Shimbo and Ishida, and utilizes the weight while the algorithm performs lookahead search. The second one, weighted update, incorporates the weight to the edges of the search graph during the learning phase. We implemented both techniques within LSS-LRTA* and evaluated them in path-planning benchmarks. We show that weighted lookahead outperforms an existing approach by Shimbo and Ishida but that it does not improve over existing approaches that do not use weights. Weighted update, on the other hand, yields performance improvements of up to one order of magnitude both in solution cost and total search time. To illustrate further the generality of weighted update, we incorporate the technique in two other well-known real-time heuristic search algorithms: LRTA*-LS and daLSS-LRTA*, and we empirically show significant improvements for LRTA*-LS and modest but still important improvements for daLSS-LRTA*. We analyze the properties of weighted update in depth, showing, among other things, that it guarantees termination. Convergence behavior of LSS-LRTA*, modified to use weighted update, is also analyzed. In such a setting, we prove solutions are w-optimal, and provide additional bounds on solution quality that in practice are tighter than w-optimality.