In this paper, we develop Fringe-Saving A* (FSA*), an incremental version of A * that repeatedly finds shortest paths in a known gridworld from a given start cell to a given goal cell while the traversability costs of cells increase or decrease. The first search of FSA * is the same as that of A*. However, FSA * is able to find shortest paths during the subsequent searches faster than A * because it reuses the beginning of the immediately preceeding A * search tree that is identical to the current A* search tree. FSA * does this by restoring the content of the OPEN list of A * at the point in time when an A * search for the current search problem could deviate from the A * search for the immediately preceeding search problem. We present first experimental results that demonstrate that FSA * can have a runtime advantage over A * and Lifelong Planning A * (LPA*), an alternative incremental version of A*. 1

Incremental search algorithms reuse information from previous searches to speed up the current search and are thus often able to find shortest paths for series of similar search problems faster than by solving each search problem independently from scratch. However, they do poorly on moving target search problems, where both the start and goal cells change over time. In this paper, we thus develop Fringe-Retrieving A * (FRA*), an incremental version of A * that repeatedly finds shortest paths for moving target search in known gridworlds. We demonstrate experimentally that it runs up to one order of magnitude faster than a variety of state-of-the-art incremental search algorithms applied to moving target search in known gridworlds.

Fringe-Saving A * is an incremental version of A * that repeatedly finds shortest paths from a fixed start cell to a fixed goal cell in a known gridworld in case the traversability of cells changes over time. It restores the content of the OPEN and CLOSED lists of A * at the point in time when an A* search for the current search problem could deviate from the A * search for the previous search problem. Thus, Fringe-Saving A * reuses the beginning of the previous A * search that is identical to the current A * search. In this paper, we generalize the correctness proof of Fringe-Saving A * to cover the case where the goal cell changes over time in addition to the traversability of cells. We then apply Fringe-Saving A * to the problem of moving an agent along a shortest path from its current cell to a fixed destination cell in a known gridworld, where the shortest path is replanned whenever the traversability of cells changes. Our experimental results show that the resulting Dynamic Fringe-Saving A * algorithm can outperform both repeated A * searches and D* Lite (a state-of-the-art incremental version of A*) in highly dynamic gridworlds, with runtime savings of up to a factor of about 2.5.

Abstract — Hybrid deliberative-reactive control architectures are a popular and effective approach to the control of robotic navigation applications. However, the design of said architectures is difficult, due to the fundamental differences in the design of the reactive and deliberative layers of the architecture. We propose a novel approach to improving system-level performance of said architectures, by improving the deliberative layer’s model of the reactive layer’s execution of its plans through the use of machine learning techniques. Quantitative and qualitative results from a physics-based simulator are presented. I.

Hybrid deliberative-reactive control architectures are a popular and effective approach to the control of robotic navigation applications. However, due to the fundamental differences in the design of the reactive and deliberative layers, the design of hybrid control architectures can pose significant difficulties. We propose a novel approach to improving system-level performance of hybrid control architectures, by incrementally improving the deliberative layer’s model of the reactive layer’s execution of its plans. Incremental supervised learning techniques are employed to learn the model. Quantitative and qualitative results from a physics-based simulator are presented.

Virtual characters in games and simulations often need to plan visually convincing paths through a crowded environment. This paper describes how crowd density information can be used to guide a large number of characters through a crowded environment. Crowd density information helps characters avoid congested routes that could lead to traffic jams. It also encourages characters to use a wide variety of routes to reach their destination. Our technique measures the desirability of a route by combining distance information with crowd density information. We start by building a navigation mesh for the walkable regions in a polygonal two-dimensional (2-D) or multilayered three-dimensional (3-D) environment. The skeleton of this navigation mesh is the medial axis. Each walkable region in the navigation mesh maintains an up-to-date density value. This density value is equal to the area occupied by all the characters inside a given region divided by the total area of this region. These density values are mapped onto the medial axis to form a weighted graph. An A search on this graph yields a backbone path for each character, and forces are used to guide the characters through the weighted environment. The characters periodically replan their routes as the density values are updated. Our experiments show that we can compute congestion-avoiding paths for tens of thousands of characters in real-time. Copyright © 2012 John Wiley & Sons, Ltd.

Incremental search algorithms, such as Generalized Fringe-Retrieving A * and D * Lite, reuse search trees from previous searches to speed up the current search and thus often find cost-minimal paths for series of similar search problems faster than by solving each search problem from scratch. However, existing incremental search algorithms have limitations. For example, D * Lite is slow on moving target search problems, where both the start and goal states can change over time. In this paper, we therefore introduce Moving Target D * Lite, an extension of D * Lite that uses the principle behind Generalized Fringe-Retrieving A * to repeatedly calculate a cost-minimal path from the hunter to the target in environments whose blockages can change over time. We demonstrate experimentally that Moving Target D * Lite is four to five times faster than Generalized Adaptive A*, which so far was believed to be the fastest incremental search algorithm for solving moving target search problems in dynamic environments.

Virtual characters in games and simulations often need to plan visually convincing paths through a crowded environment. This paper describes how crowd density information can be used to guide a large number of characters through a crowded environment. Crowd density information helps characters avoid congested routes that could lead to traffic jams. It also encourages characters to use a wide variety of routes to reach their destination. Our technique measures the desirability of a route by combining distance information with crowd density information. We start by building a navigation mesh for the walkable regions in a polygonal 2D or multi-layered 3D environment. The skeleton of this navigation mesh is the medial axis. Each walkable region in the navigation mesh maintains an up-to-date density value, given by the fraction of the area that is being occupied by characters. These density values are mapped onto the medial axis to form a weighted graph. An A * search on this graph yields a backbone path for each character, and forces are used to guide the characters through the weighted environment. The characters periodically replan their routes as the density values are updated. Our experiments show that we can compute congestionavoiding paths for tens of thousands of characters in real-time. 1

Abstract—Several disciplines, including artificial intelligence, operations research and many others, study how to make good decisions. In this overview article, we argue that the key to making progress in our research area is to combine their ideas, which often requires serious technical advances to reconcile their different assumptions and methods in a way that results in synergy among them. To illustrate this point, we give a broad overview of our ongoing research on search and planning (with a large number of students and colleagues, both at the University of Southern California and elsewhere) to demonstrate how to combine ideas from different decision making disciplines. For example, we describe how to combine ideas from artificial intelligence, operations research, and utility theory to create the foundations for building decision support systems that fit the risk preferences of human decision makers in high-stake one-shot decision situations better than current systems. We also describe how to combine ideas from artificial intelligence, economics, theoretical computer science and operations research to build teams of robots that use auctions to distribute tasks autonomously among themselves, and give several more examples. Index Terms—agents, ant robotics, artificial intelligence, auction-based coordination, decision theory, dynamic programming, economics, freespace assumption, goal-directed navigation, greedy online planning, heuristic search, high-stake one-shot decision making, incremental heuristic search, Markov decision processes, multi-agent systems, nonlinear utility functions, operations research, planning, real-time heuristic search, reinforcement learning, risk preferences, robotics, scarce resources, sequentialsingle item auctions, terrain coverage, utility theory. I.