Probabilistic Localization Algorithm Research Articles

Smartphone-based localization for blind navigation in building-scale indoor environments

Continuous, accurate, and real-time smartphone-based localization is a promising technology for supporting independent mobility of people with visual impairments. However, despite extensive research on indoor localization techniques, localization technologies are still not ready for deployment in large and complex environments such as shopping malls and hospitals, where navigation assistance is needed most. We identify six key challenges for accurate smartphone localization related to the large-scale nature of the navigation environments and the user’s mobility. To address these challenges, we present a series of techniques that enhance a probabilistic localization algorithm. The algorithm utilizes mobile device inertial sensors and Received Signal Strength (RSS) from Bluetooth Low Energy (BLE) beacons. We evaluate the proposed system in a 21,000 m2 shopping mall that includes three multi-story buildings and a large open underground passageway. Experiments conducted in this environment demonstrate the effectiveness of the proposed technologies to improve localization accuracy. Field experiments with visually impaired participants confirm the practical performance of the proposed system in realistic use cases.

Robot navigation in orchards with localization based on Particle filter and Kalman filter

Fruit production in orchards currently relies on high labor inputs. Concerns arising from the increasing labor cost and shortage of labor can be mitigated by the availability of an autonomous orchard robot. A core feature for every mobile orchard robot is autonomous navigation, which depends on sensor-based robot localization in the orchard environment. This research validated the applicability of two probabilistic localization algorithms that used a 2D LIDAR scanner for in-row robot navigation in orchards. The first localization algorithm was a Particle filter (PF) with a laser beam model, and the second was a Kalman filter (KF) with a line-detection algorithm. We evaluated the performance of the two algorithms when autonomously navigating a robot in a commercial Dutch apple orchard. Two experiments were executed to assess the navigation performance of the two algorithms under comparable conditions. The first experiment assessed the navigation accuracy, whereas the second experiment tested the algorithms’ robustness. In the first experiment, when the robot was driven with 0.25 m/s the root mean square error (RMSE) of the lateral deviation was 0.055 m with the PF algorithm and 0.087 m with the KF algorithm. At 0.50 m/s, the RMSE was 0.062 m with the PF algorithm and 0.091 m with the KF algorithm. In addition, with the PF the lateral deviations were equally distributed to both sides of the optimal navigation line, whereas with the KF the robot tended to navigate to the left of the optimal line. The second experiment tested the algorithms’ robustness to cope with missing trees in six different tree row patterns. The PF had a lower RMSE of the lateral deviation in five tree patterns. In three out of the six patterns, navigation with the KF led to lateral deviations that were biased to the left of the optimal line. The angular deviations of the PF and the KF were in the same range in both experiments. From the results, we conclude that a PF with laser beam model is to be preferred over a line-based KF for the in-row navigation of an autonomous orchard robot.

DPC-Net: Deep Pose Correction for Visual Localization

We present a novel method to fuse the power of deep networks with the computational efficiency of geometric and probabilistic localization algorithms. In contrast to other methods that completely replace a classical visual estimator with a deep network, we propose an approach that uses a convolutional neural network to learn difficult-to-model corrections to the estimator from ground-truth training data. To this end, we derive a novel loss function for learning SE(3) corrections based on a matrix Lie groups approach, with a natural formulation for balancing translation and rotation errors. We use this loss to train a Deep Pose Correction network (DPC-Net) that predicts corrections for a particular estimator, sensor and environment. Using the KITTI odometry dataset, we demonstrate significant improvements to the accuracy of a computationally-efficient sparse stereo visual odometry pipeline, that render it as accurate as a modern computationally-intensive dense estimator. Further, we show how DPC-Net can be used to mitigate the effect of poorly calibrated lens distortion parameters.

Localization of an Autonomous Rail-Guided Robot Using Particle Filter

This work presents a probabilistic localization algorithm for the autonomous navigation system of a rail-guided robot, named DORIS, aimed at performing inspection and monitoring tasks in offshore facilities. A particle filter, which integrates odometry with inertial measurements, laser scans and image data, is used to estimate the robot localization on the rail. A novel technique, based on the recent history of events observed by the robot, is proposed to enhance the standard Monte Carlo localization algorithm, improving the robustness and convergence of the estimation and reducing its computational complexity. Simulations with experimental data acquired in field tests show that the proposed algorithm estimates the robot position with an absolute error smaller than 25cm (0.2% of the total rail length). The algorithm is able to deal with important localization issues of autonomous robots, which are the global localization and the kidnapped robot problems, proving to be better than odometry or a standard particle filter.

Tracking porpoise underwater movements in tidal rapids using drifting hydrophone arrays

The growing interest in generating electrical power from tidal currents using tidal turbine generators raises a number of environmental concerns, including the risk that cetaceans might be injured or killed through collision with rotating turbine blades. To understand this risk we need better information on how cetaceans use tidal rapid habitats and in particular their underwater movements and dive behavior. Focusing on harbor porpoises, a European protected species, we have developed an approach which uses time of arrival differences of narrow band high frequency (NBHF) clicks detected on large aperture hydrophone arrays drifting in tidal rapids, to determine dive tracks of porpoises underwater. Probabilistic localization algorithms have been developed to filter echoes and provide accurate 2D or geo-referenced 3D locations. Calibration trials have been carried out that show that the system can provide depth and location data with submeter errors. Data collected over three seasons in tidal races around Scotland has provided new insights into how harbor porpoises are using these unique habitats, information vital for assessing the risk tidal turbines may pose.

Heterogeneous Cooperative Localization for Social Networks

Location-aware techniques has become a hot research topic with great value in commercial and military applications. Cooperative localization, which utilizes multiple sensors in portable devices to estimate locations of the mobile users in the social networks, is one of the most promising solution for the indoor geo-location. Traditional cooperative localization methods are based on ranging techniques, they are highly dependent on the distance interpreted from the received signal strength (RSS) or time of arrival from anchors. However, a precise ranging procedure demands high performance hardware which would increase the cost to the current mobile platform. In this paper, we describes four ranging-free probabilistic cooperative localization algorithms: centroid scheme, nearest neighbor scheme, kernel scheme and AP density scheme to improve the accuracy for the indoor geo-location using current mobile devices. Since the GPS sensor embedded in the smart phone is able to provide accurate location information in the outdoor area, those mobile nodes can be used as calibrated anchors. The position of the indoor mobile node can be estimated by exchanging locations and RSSs from shared wireless access points information between the target node and anchor nodes. An empirical evaluation of the system is given to demonstrate the feasibility of these cooperative localization algorithms by reporting the results in a real-world environments, e.g. suburban area and city downtown. Moreover, we compared our results with the WiFi positioning system made by Skyhook Wireless to validate the accuracy of the proposed algorithms. Meanwhile, a Monte Carlo simulation is carried out to evaluate the performance of the cooperative algorithms under different scenarios. Results show that given the same scenario setting, the AP density scheme and kernel scheme outperform than other schemes.

Localization Algorithms for Wireless Sensor Retrieval

In wireless sensor networks (WSNs), localization has many important applications, among which wireless sensor retrieval bears special importance for cost saving, data analysis and security purposes. Localization for sensor retrieval is especially challenging due to the fact that the number and locations of these sensors are both unknown. In this paper, we propose two probabilistic localization algorithms that iteratively identify the locations of multiple wireless sensors in WSNs, one of which calculates location information offline, and the other online. In both algorithms, we implement a two-step localization process — the first step is called Grid-LEGMM (grid location estimation based on the Gaussian mixture model), a coarse-grain location search using grids by choosing the proper number and locations of the wireless sensors that maximize a likelihood estimation, and the second step is called EM-LEGMM (expectation maximization based on the Gaussian mixture model), which uses the EM-method to refine the results of Grid-LEGMM. An additional step in the online localization algorithm is a credit-based filtering mechanism that removes spurious sensor locations. The performance of both offline and online localization algorithms are analyzed using the Cramer–Rao lower bound (CRLB), and evaluated using simulations and real testbed experiments.

Grid division-based probabilistic localization algorithm for WSN

The Received Signal Strength Indicator(RSSI)-based nodes localization algorithms are extensively applied in wireless sensor networks.Due to the influence of wireless signal attenuation and the channel barrier shielding,the localization accuracy of these algorithms are not high.A grid division-based probabilistic localization algorithm was put forward.This new algorithm first divided the area where the unknown nodes may exist into a number of grids.Then,it calculated out the confidence of every grid under the assumption that the unknown node was just in that grid.Finally,it calculated out the coordinates of the unknown node based on several geometric centers of the grids of high confidence.The localization result of this algorithm is less influenced by the noise pollution and is of high position accuracy.

Robust Monte Carlo localization for mobile robots

Mobile robot localization is the problem of determining a robot's pose from sensor data. This article presents a family of probabilistic localization algorithms known as Monte Carlo Localization (MCL). MCL algorithms represent a robot's belief by a set of weighted hypotheses (samples), which approximate the posterior under a common Bayesian formulation of the localization problem. Building on the basic MCL algorithm, this article develops a more robust algorithm called Mixture-MCL, which integrates two complimentary ways of generating samples in the estimation. To apply this algorithm to mobile robots equipped with range finders, a kernel density tree is learned that permits fast sampling. Systematic empirical results illustrate the robustness and computational efficiency of the approach.