Andrew Tridgell

Learning to Play Chess Using Temporal Differences

In this paper we present TDLEAF(λ), a variation on the TD(λ) algorithm that enables it to be used in conjunction with game-tree search. We present some experiments in which our chess program “KnightCap” used TDLEAF(λ) to learn its evaluation function while playing on Internet chess servers. The main success we report is that KnightCap improved from a 1650 rating to a 2150 rating in just 308 games and 3 days of play. As a reference, a rating of 1650 corresponds to about level B human play (on a scale from E (1000) to A (1800)), while 2150 is human master level. We discuss some of the reasons for this success, principle among them being the use of on-line, rather than self-play. We also investigate whether TDLEAF(λ) can yield better results in the domain of backgammon, where TD(λ) has previously yielded striking success.

Interactive simulation of bushfires in heterogeneous fuels

The program IGNITE, developed by the authors, is a landscape fire modelling system that deals with fires in heterogeneous fuels. Landscapes are represented as cellular automata (grids of pixels) and fire spread is modelled as an epidemic process. An integrated geographic information system permits the importing and editing of maps from compatible sources, such as satellite imagery. Maps, models and other information are organized as scenarios; historical fires can be recorded and replayed. Modules are being developed for application to fire prevention, fire suppression, land-use management, and to training and education. An illustration of using the system to deal with heterogeneous fuel is its application to the problem of percolation in patchy fuel.

Implementing MPI under AP/Linux

A preliminary MPI library has been implemented for the Fujitsu AP1000+ multicomputer running the AP/Linux operating system. Under this environment, parallel programs may be dedicated to a fixed partition, or a number of parallel programs may share a partition. Therefore, the MPI library has been constructed so that messaging operations can be driven by polling and/or interrupt techniques. It has been found that polling works well when a single parallel program is running on a given partition, and that interrupt-driven communication makes far better use of the machine when multiple parallel programs are executing. Gang scheduling of multiple parallel programs which use polling was found to be relatively ineffective.

A General-Purpose Parallel Sorting Algorithm

A parallel sorting algorithm is presented for general purpose internal sorting on MIMD machines. The algorithm initially sorts the elements within each node using a serial sorting algorithm, then proceeds with a two-phase parallel merge. The algorithm is comparison-based and requires additional storage of order the square root of the number of elements in each node. Performance of the algorithm on the Fujitsu AP1000 MIMD supercomputer is discussed.

A fast, storage-efficient parallel sorting algorithm

A parallel sorting algorithm is presented for storage-efficient internal sorting on MIMD machines. The algorithm first sorts the elements within each node using a serial based algorithm, then a two-phase parallel merge. It requires additional storage of order of the square root of the number of elements in each node. Performance of the algorithm on two general-purpose MIMD machines, the Fujitsu AP1000 and the Thinking Machines CM5, is examined. The algorithm is suitable for implementation on special-purpose parallel machines, e.g., parallel database machines.<<ETX>>

Low-cost differential GPS for field robotics

In this paper we detail the design and implementation of a Differential GPS system that uses only low-cost, light-weight and widely available GPS modules. Such a system is desirable for modern field robotics in order to improve the performance of mapping, navigation, collision avoidance and other common tasks, especially on payload-constrained robots. The particular noise and error modes present in the low-cost system are investigated and algorithms introduced to minimise their effect. The system is demonstrated in the context of field robotics for stationary, ground-based and airbourne agents. It is found that the majority of the error due to the low cost components can be manipulated to be of a form easily mitigated for most robots. The system is shown to provide good results in the horizontal plane in open environments such as those typically used for outdoor experimental robotics while being one to two orders of magnitude cheaper than comparable commercial systems.

Architecture and implementation of an affordable differential GPS system

In this paper we detail the implementation of a Differential GPS system using only low-cost GPS receivers. The nature of the errors generated by low-cost modules is investigated and contrasted against typical error models of traditional, commercial DGPS systems. It is found that the error profile unique to low-cost GPS modules leads to a system that offers particular advantage in the horizontal plane compared to uncorrected receivers and as such is especially suited for tasks such as mapping and field robotics. A particular incarnation is demonstrated on a mobile robot with experimental results from static and moving tests showing an accuracy in the plane on the order of one metre.