Change detection is important for autonomous perception systems that operate in dynamic environments. Mapping and tracking components commonly handle two ends of the dynamic spectrum: stationarity and rapid motion. This paper presents a fast algorithm for 3D change detection from LIDAR or equivalent optical range sensors, that can operate from arbitrary viewpoints and can detect fast and slow dynamics. Distinct from prior work, the method explicitly detects changes in the world, and suppresses apparent changes in the data due to exploration at frontiers or behind occlusions. Comprehensive experimentation is performed to assess the performance in several application domains. Sample data and source code are provided.

This paper presents a method for pairwise 3D alignment which solves data association by matching scan segments across scans. Generating accurate segment associations allows to run a modified version of the Iterative Closest Point (ICP) algorithm where the search for point-to-point correspondences is constrained to associated segments. The novelty of the proposed approach is in the segment matching process which takes into account the proximity of segments, their shape, and the consistency of their relative locations in each scan. Scan segmentation is here assumed to be given (recent studies provide various alternatives [10], [19]). The method is tested on seven sequences of Velodyne scans acquired in urban environments. Unlike various other standard versions of ICP, which fail to recover correct alignment when the displacement between scans increases, the proposed method is shown to be robust to displacements of several meters. In addition, it is shown to lead to savings in computational times which are potentially critical in real-time applications.

This paper examines the notions of consistency and conservativeness for data fusion involving dependent information, where the degree of dependency is unknown. We consider these notions in a general sense, for non-Gaussian probability distributions, in terms of structural consistency and information processing, in particular the counting of common information. We consider the role of entropy in defining a conservative fusion rule. Finally, we investigate the geometric mean density (GMD) as a particular fusion rule, which generalises the Covariance Intersection rule to non-Gaussian pdfs. We derive key properties to demonstrate that the GMD is both conservative and effective in combining information from dependent sources.

This paper presents a probabilistic framework for road geometry estimation using a millimetre wave radar. It aims at estimating the geometry of roads without assuming any particular infrastructure such as lane marks. It provides also the vehicle location with respect to the edges of the road. This system employs a radar sensor in view of its robustness to weather conditions such as fog, dust, rain and snow. The proposed approach is robust to noisy measurements since the radar target locations are modelled as Gaussian distributions. These observations are integrated into a Kalman Particle filter to estimate the posterior distribution of the parameters that best describe the geometry of the road. Experimental results using data acquired on a highway road are presented. The effectiveness of the proposed approach is demonstrated by a qualitative analysis of the results.

The problem of cooperative navigation for a team of platforms employing inter-platform observations is investigated. A decentralised solution in the framework of an information filter with delayed states is presented. In this structure, each platform first estimates its motion using only local sensor data, then shares its information across the network using an algorithm that employs a distributed Cholesky modification. The decentralised solution permits each platform to act in the same modular manner, providing robustness to individual platform failure. The solution yields linear minimum mean-square error estimation performance. As such the estimates generated are optimal; it generates exactly the same estimates as would a conventional extended Kalman filter (EKF), if given the same data. Efficient sparse implementation is accomplished without resorting to approximate methods. Simulation experiments employing a team of ten mobile platforms are described and used to evaluate the decentralised estimation performance. The robustness, flexibility, and cost of the decentralised approach are analyzed and compared with an existing distributed solution.

In this paper we address the problem of closing the loop from perception to action selection for unmanned ground vehicles, with a focus on navigating slopes. A new non-parametric learning technique is presented to generate a mobility representation where the maximum feasible speed is used as a criterion to classify the world. The inputs to the algorithm are terrain gradients derived from an elevation map and past observations of wheel slip. It is argued that such a representation can aid in path planning with improved selection of vehicle heading and velocity in off-road slopes. In addition, an information theoretic test is proposed to validate a chosen proprioceptive representation (such as slip) for mobility map generation. Results of mobility map generation and its benefits to path planning are shown.

Novel lazy Lauritzen-Spiegelhalter (LS), lazy Hugin and lazy Shafer-Shenoy (SS) algorithms are devised for Gaussian Bayesian networks (BNs). In the lazy algorithms, the clique potentials and separator potentials are kept in combinable decomposed form instead of combined to be a single valuation in conventional junction tree algorithms. By employing decomposed form potentials, the independence relations between variables are explored online and the directed graph information is utilized in the message calculations. In the proposed algorithms, a consistent junction tree with the evidence entered can be obtained by a single round of message passing. The moments form parametrization of Gaussian distributions allows the deterministic relationships between variables. Preliminary analysis shows that the lazy LS algorithm and the lazy Hugin algorithm are more computationally efficient than the lazy SS algorithm, especially when there are multiple items of evidence to be incorporated.

This paper presents HybridSLAM: an approach to SLAM which combines the strengths and avoids the weaknesses of two popular mapping strategies: FastSLAM and EKF-SLAM. FastSLAM is used as a front-end, producing local maps which are periodically fused into an EKF-SLAM back-end. The use of FastSLAM locally avoids linearisation of the vehicle model and provides a high level of robustness to clutter and ambiguous data association. The use of EKFSLAM globally allows uncertainty to be remembered over long vehicle trajectories, avoiding FastSLAM’s tendency to become over-confident. Extensive trials in randomly-generated simulated environments show that HybridSLAM significantly out-performs either pure approach. The advantages of HybridSLAM are most pronounced in cluttered environments where either pure approach encounters serious difficulty. In addition, the HybridSLAM algorithm is experimentally validated in a real urban environment.

The paper investigates a technique for computing conservative data fusion for Gaussian mixture model (GMM) in decentralized networks with any topology. The main advantage of conservative solutions is that they do not deteriorate the performance of a sensor network in presence of any kind of correlations. The paper exploits normalize geometric mean for computing conservative data fusion. It computes normalized geometric mean by Newton generalized binomial theorem and Monte Carlo technique. It is shown that the solution by Newton's generalized binomial theorem exhibits divergence and numerical instability. On the other hand, Monte Carlo technique offers conservative solution. The tradeoffs are that it requires considerable computational time and is expensive as large numbers of samples are required to get statistical accuracy.

This paper presents algorithms for consistent joint localisation and tracking of multiple targets in wireless sensor networks under the decentralised data fusion (DDF) paradigm where particle representations of the state posteriors are communicated. This work differs from previous work as more generalised methods have been developed to account for correlated estimation errors that arise due to common past information between two discrete particle sets. The particle sets are converted to continuous distributions for communication and inter-nodal fusion. Common past information is then removed by a division operation of two estimates so that only new information is updated at the node. In previous work, the continuous distribution used was limited to a Gaussian kernel function. This new method is compared to the optimal centralised solution where each node sends all observation information to a central fusion node when received. Results presented include a real-time application of the DDF operation of division on data logged from field trials.

Scan-SLAM is a simultaneous localisation and mapping algorithm that combines scan-matching methods with recursive estimation of landmark locations (using an EKF or other Bayesian filter). The scan-matching capability allows landmarks with arbitrary shapes to be modelled directly by sensed data and tracked within a conventional filter framework. This paper presents the essential Scan-SLAM algorithm, and implementation details for application with a scanning range-laser: segmentation, alignment, covariance estimation, data association, and landmark model augmentation.

This paper discusses a method that enables reliable docking of a mobile robot with a rectangular container. The process involves an aligned a oach to the container while avoiding any obstacles in the region and avoiding collision with the container itself. A modular behaviour-based architecture called the Distributed Architecture for Mobile Navigation (DAMN) is used. Raw sensor data is processed to produce robust sensors that provide input data to the behaviour modules. Centralised arbiters then asynchronously process the behaviour outputs and detennine the set points for the mobile robot drive and steering actuators. Testing of the virtual :sensors and behaviour-based algorithms was perfqrmed on an indoor mobile robot, SydNav, with wheel-encoders and a scanning range laser as its sensors. Further testing of a container poseestimation sensor took place on a quayside cargohandling vehicle (a straddle-carrier); again using scanning lasers for its sensorial input.

Decentralised estimation of heterogeneous sensors is performed on an outdoor network. Attributes such as position, appearance, and identity represented by non-Gaussian distributions are used in in the fusion process. It is shown here that real-time decentralised data fusion of non-Gaussian estimates can be used to build rich environmental maps. Human operators are also used as additional sensors in the network to complement robotic information.

Delayed-state decentralised data fusion (DS-DDF) is proposed as a general methodology for consistent DDF, which does not impose constraints on network topology. The resulting estimates, although lagged in time, are optimal, equal to a centralised solution. The method is demonstrated in the context of dynamic node tracking and localisation, where a team of mobile robots track each other’s position to obtain a joint estimate of the position of every team member.

Although full 3D navigation and mapping is recognized as one of the most important challenges for autonomous navigation, the lack of robust sensors, providing 3D information in real time, has burdened the progress in this direction. This paper presents our ongoing work towards the deployment of an integrated sensing system for 3D mapping in outdoor environments. We first describe a 3D data acquisition architecture based on a standard 2D laser. Techniques for registering scans using a scan matching procedure and for estimating the errors are then introduced. We finally present results showing the performance of the proposed architecture in real outdoor environments by means of the integration of the 3D scans with dead reckoning and inertial measurement unit (IMU) information

This paper presents two solutions for performing decentralised particle filtering in view of non-linear, non-Gaussian tracking in sensor networks. The issue is that no known methods exist to deal with correlated estimation errors due to common past information between two discrete particle sets. The first method transforms the particles to a Gaussian mixture model, the second approximates the set by a Parzen density estimate. Both of these representations accommodate consistent fusion and maintain accurate summaries of the particles. Requiring less bandwidth than particle representations, transformations to GMMs or Parzen representations for communication provide an added advantage. The accuracy in which the algorithms summarise the particle set, fusion methods and bandwidth requirements of each representation will be compared. Our results show that whilst less GMM components are required to summarise the sample statistics, the decentralised fusion solution using Parzen representations yields a more accurate result

This paper discusses the recursive Bayesian formulation of the simultaneous localization and mapping (SLAM) problem in which probability distributions or estimates of absolute or relative locations of landmarks and vehicle pose are obtained. The paper focuses on three key areas: computational complexity; data association; and environment representation.