Sandip Aine

Adaptive parameter control of evolutionary algorithms to improve quality-time trade-off

Parameter control of evolutionary algorithms (EAs) poses special challenges as EA uses a population and requires many parameters to be controlled for an effective search. Quality improvement is dependent on several factors, such as, fitness estimation, population diversity and convergence rate. A widely practiced approach to identify a good set of parameters for a particular class of problem is through experimentation. Ideally, the parameter selection should depend on the resource availability, and thus, a rigid choice may not be suitable. In this work, we propose an automated framework for parameter selection, which can adapt according to the constraints specified. To condition the parameter choice through resource constraint/utilization, we consider two typical scenarios, one where maximum available run-time is pre-specified and the other in which a utility function modeling the quality-time trade-off is used instead of a rigid deadline. We present static and dynamic parameter selection strategies based on a probabilistic profiling method. Experiments performed with traveling salesman problem (TSP) and standard cell placement problem show that an informed adaptive parameter control mechanism can yield better results than a static selection.

Multi-Heuristic A*

The performance of heuristic search-based planners depends heavily on the quality of the heuristic function used to focus the search. These algorithms work fast and generate high-quality solutions, even for high-dimensional problems, as long as they are given a well-designed heuristic function. On the other hand, their performance can degrade considerably if there are large heuristic depression regions, i.e. regions in the search space where heuristic values do not correlate well with the actual cost-to-goal values. Consequently, the research in developing an efficient planner for a specific domain becomes the design of a good heuristic function. However, for many domains, it is hard to design a single heuristic function that captures all of the complexities of the problem. Furthermore, it is hard to ensure that heuristics are admissible (provide lower bounds on the cost-to-goal) and consistent, which is necessary for A* like searches to provide guarantees on completeness and bounds on sub-optimality. In this paper, we develop a novel heuristic search, called Multi-Heuristic A* (MHA*), that takes in multiple, arbitrarily inadmissible heuristic functions in addition to a single consistent heuristic, and uses all of them simultaneously to search in a way that preserves guarantees on completeness and bounds on sub-optimality. This enables the search to combine very effectively the guiding powers of different heuristic functions and simplifies dramatically the process of designing heuristic functions by a user because these functions no longer need to be admissible or consistent. We support these claims with experimental analysis on several domains ranging from inherently continuous domains such as full-body manipulation and navigation to inherently discrete domains such as the sliding tile puzzle.

Anytime column search

Anytime heuristic search algorithms are widely applied where best-first search algorithms such as A* require large or often unacceptable amounts of time and memory. Anytime algorithms produce a solution quickly and iteratively improve the solution quality. In this paper, we propose novel anytime heuristic search algorithms with a common underlying strategy called Column Search. The proposed algorithms are complete and guarantee to produce an optimal solution. Experimental results on sliding-tile puzzle problem, traveling salesman problem, and robotic arm trajectory planning problem show the efficacy of proposed methods compared to state-of-the-art anytime heuristic search algorithms.

MAWA∗-a memory-bounded anytime heuristic-search algorithm.

This paper presents a heuristic-search algorithm called Memory-bounded Anytime Window A*(MAWA*), which is complete, anytime, and memory bounded. MAWA* uses the window-bounded anytime-search methodology of AWA* as the basic framework and combines it with the memory-bounded A* -like approach to handle restricted memory situations. Simple and efficient versions of MAWA* targeted for tree search have also been presented. Experimental results of the sliding-tile puzzle problem and the traveling-salesman problem show the significant advantages of the proposed algorithm over existing methods.

HEURISTIC SEARCH UNDER CONTRACT

In this article, we present a heuristic search technique (Contract Search) that can be adapted automatically for a specific node contract. We analyze the node expansion characteristics of best‐first search techniques and identify a probabilistic model (rank profiles) that characterizes the search under restricted expansions. We use the model to formulate an optimal strategy to choose level dependent restriction bounds, maximizing the probability of obtaining the optimal cost goal node under the specified contract. We analyze the basic properties of the rank profiles and establish its relation with the search space configuration and heuristic error distributions. We suggest an approximation scheme for the profile function for unknown search spaces. We show how the basic framework can be adapted to achieve different objectives (like optimizing the expected quality) considering multiple goals and approximate solutions. Experimental comparison with anytime search techniques like ARA* and beam search on a number of search problems shows that Contract Search outperforms these techniques over a range of contract specifications.

Truncated incremental search

Incremental heuristic search algorithms reuse their previous search efforts whenever these are available. As a result, they can often solve a sequence of similar planning problems faster than planning from scratch. State-of-the-art incremental heuristic searches (such as LPA*, D* and D* Lite) work by propagating cost changes to all the states in the search tree whose g values (the costs of computed paths from the start state) are no longer optimal. This work is based on the observation that while a complete propagation of cost changes is essential to ensure optimality, the propagations can be stopped earlier if we are looking close-to-optimal solutions instead of the optimal one. We develop a framework called Truncated Incremental Search that builds on this observation and uses a target suboptimality bound to efficiently restrict cost propagations. We present two truncation based algorithms, Truncated LPA* (TLPA*) and Truncated D* Lite (TD* Lite), for bounded suboptimal planning and navigation in dynamic graphs. We also develop an anytime replanning algorithm, Anytime Truncated D* (ATD*), that combines the inflated heuristic search with truncation, in an anytime manner. We discuss the theoretical properties of these algorithms proving their correctness and efficiency, and present experimental results on 2D and 3D (x, y, heading) path planning domains evaluating their performance. The empirical results show that the truncated incremental searches can provide significant improvement in runtime over existing incremental search algorithms, especially when searching for close-to-optimal solutions in large, dynamic graphs.

Incremental Beam search

Beam search is a heuristic search algorithm that explores a state-space graph by expanding w most promising nodes at each level (depth) of the graph, where w is called the beam-width which is taken as input from the user. The quality of the solution produced by beam search does not always monotonically improve with the increase in beam-width making it difficult to choose an appropriate beam-width for effective use. We present an algorithm called Incremental Beam Search (IncB) which guarantees monotonicity, and is also anytime in nature. Experimental results on the sliding-tile puzzle, the traveling salesman, and the single-machine scheduling problems show that IncB significantly outperforms basic monotonic methods such as iterative widening beam search as well as some of the state-of-the-art anytime heuristic search algorithms in terms of the quality of the solution produced at the end as well as the anytime performance.

Anytime pack search

Abstract Heuristic search is one of the fundamental problem solving techniques in artificial intelligence, which is used in general to efficiently solve computationally hard problems in various domains, especially in planning and optimization. In this paper, we present an anytime heuristic search algorithm called anytime pack search (APS) which produces good quality solutions quickly and improves upon them over time, by focusing the exploration on a limited set of most promising nodes in each iteration. We discuss the theoretical properties of APS and show that it is complete. We also present the complexity analysis of the proposed algorithm on a tree state-space model and show that it is asymptotically of the same order as that of A*, which is a widely applied best-first search method. Furthermore, we present a parallel formulation of the proposed algorithm, called parallel anytime pack search (PAPS), which is applicable for searching tree state-spaces. We theoretically prove the completeness of PAPS. Experimental results on the sliding-tile puzzle problem, traveling salesperson problem, and single machine scheduling problem depict that the proposed sequential algorithm produces much better anytime performance when compared to some of the existing methods. Also, the proposed parallel formulation achieves super-linear speedups over the sequential method.

An Automated Meta-Level Control Framework for Optimizing the Quality–Time Tradeoff of VLSI Algorithms

We address the problem of optimizing the quality-time tradeoff of very large scale integration (VLSI) computer-aided design (CAD) algorithms working under various constrained environments. We present a unified meta-reasoning framework to automatically control the progress of iteratively improving CAD algorithms. The control framework uses profile knowledge about the quality-time relation of the algorithms used and generates a combined strategy for time allocation and parameter control that optimizes the expected tradeoff. We present specific formulations for handling single and multiple problems (both dependent and independent) under various constraints. We use the proposed strategies for adjusting the control parameters of standard simulated annealing and genetic algorithm based techniques used in VLSI optimization along with an appropriate time allocation suited for different constraint specifications. Application on several classical problems in the VLSI domain shows that significant improvement in quality-time tradeoff can be achieved.

Anytime Pack Heuristic Search

Heuristic search is a fundamental problem solving technique in artificial intelligence. In this paper, we propose an anytime heuristic search algorithm called Anytime Pack Search (APS) which helps in solving hard combinatorial search problems efficiently. It expands nodes of a search graph in a localized best-first manner so as to converge towards good quality solutions at regular intervals. APS is complete on bounded graphs and guarantees termination with an optimal solution. Experimental results on the sliding-tile puzzle problem, the traveling salesman problem, and the single-machine scheduling problem show that APS significantly outperforms some of the state-of-the-art anytime algorithms.