Model-Based Slippage Estimation to Enhance Planetary Rover Localization with Wheel Odometry

The exploration of planetary surfaces with unmanned wheeled vehicles will require sophisticated software for guidance, navigation and control. Future missions will be designed to study harsh environments that are characterized by rough terrains and extreme conditions. An accurate knowledge of the trajectory of planetary rovers is fundamental to accomplish the scientific goals of these missions. This paper presents a method to improve rover localization through the processing of wheel odometry (WO) and inertial measurement unit (IMU) data only. By accurately defining the dynamic model of both a rover’s wheels and the terrain, we provide a model-based estimate of the wheel slippage to correct the WO measurements. Numerical simulations are carried out to better understand the evolution of the rover’s trajectory across different terrain types and to determine the benefits of the proposed WO correction method.

Calibration and Improvement of an Odometry Model with Dynamic Wheel and Lateral Dynamics Integration

Localization is a key part of an autonomous system, such as a self-driving car. The main sensor for the task is the GNSS, however its limitations can be eliminated only by integrating other methods, for example wheel odometry, which requires a well-calibrated model. This paper proposes a novel wheel odometry model and its calibration. The parameters of the nonlinear dynamic system are estimated with Gauss-Newton regression. Due to only automotive-grade sensors are applied to reach a cost-effective system, the measurement uncertainty highly corrupts the estimation accuracy. The problem is handled with a unique Kalman-filter addition to the iterative loop. The experimental results illustrate that without the proposed improvements, in particular the dynamic wheel assumption and integrated filtering, the model cannot be calibrated precisely. With the well-calibrated odometry, the localization accuracy improves significantly and the system can be used as a cost-effective motion estimation sensor in autonomous functions.

A Novel IMU Extrinsic Calibration Method for Mass Production Land Vehicles

Multi-modal sensor fusion has become ubiquitous in the field of vehicle motion estimation. Achieving a consistent sensor fusion in such a set-up demands the precise knowledge of the misalignments between the coordinate systems in which the different information sources are expressed. In ego-motion estimation, even sub-degree misalignment errors lead to serious performance degradation. The present work addresses the extrinsic calibration of a land vehicle equipped with standard production car sensors and an automotive-grade inertial measurement unit (IMU). Specifically, the article presents a method for the estimation of the misalignment between the IMU and vehicle coordinate systems, while considering the IMU biases. The estimation problem is treated as a joint state and parameter estimation problem, and solved using an adaptive estimator that relies on the IMU measurements, a dynamic single-track model as well as the suspension and odometry systems. Additionally, we show that the validity of the misalignment estimates can be assessed by identifying the misalignment between a high-precision INS/GNSS and the IMU and vehicle coordinate systems. The effectiveness of the proposed calibration procedure is demonstrated using real sensor data. The results show that estimation accuracies below 0.1 degrees can be achieved in spite of moderate variations in the manoeuvre execution.

Least Squares Optimization: From Theory to Practice

Nowadays, Nonlinear Least-Squares embodies the foundation of many Robotics and Computer Vision systems. The research community deeply investigated this topic in the last few years, and this resulted in the development of several open-source solvers to approach constantly increasing classes of problems. In this work, we propose a unified methodology to design and develop efficient Least-Squares Optimization algorithms, focusing on the structures and patterns of each specific domain. Furthermore, we present a novel open-source optimization system that addresses problems transparently with a different structure and designed to be easy to extend. The system is written in modern C++ and runs efficiently on embedded systemsWe validated our approach by conducting comparative experiments on several problems using standard datasets. The results show that our system achieves state-of-the-art performances in all tested scenarios.

A Semantic-Based Gas Source Localization with a Mobile Robot Combining Vision and Chemical Sensing

This paper addresses the localization of a gas emission source within a real-world human environment with a mobile robot. Our approach is based on an efficient and coherent system that fuses different sensor modalities (i.e., vision and chemical sensing) to exploit, for the first time, the semantic relationships among the detected gases and the objects visually recognized in the environment. This novel approach allows the robot to focus the search on a finite set of potential gas source candidates (dynamically updated as the robot operates), while accounting for the non-negligible uncertainties in the object recognition and gas classification tasks involved in the process. This approach is particularly interesting for structured indoor environments containing multiple obstacles and objects, enabling the inference of the relations between objects and between objects and gases. A probabilistic Bayesian framework is proposed to handle all these uncertainties and semantic relations, providing an ordered list of candidates to be the source. This candidate list is updated dynamically upon new sensor measurements to account for objects not previously considered in the search process. The exploitation of such probabilities together with information such as the locations of the objects, or the time needed to validate whether a given candidate is truly releasing gases, is delegated to a path planning algorithm based on Markov decision processes to minimize the search time. The system was tested in an office-like scenario, both with simulated and real experiments, to enable the comparison of different path planning strategies and to validate its efficiency under real-world conditions.

Surface Estimation ICP Algorithm for Building a 3D Map by a Scanning LRF

A new three-dimensional (3D) map building method based on Laser Range Finder (LRF) has been proposed in this research, performing a surface estimation with the Iterative Closest Point (ICP) algorithm. While a mobile robot is navigating in an unknown environment, the entire environment cannot be scanned by LRF since kinematic features of the mobile robot and surface conditions are dynamically changing. To resolve this difficulty in building a 3D map while the mobile robot is navigating, a surface estimation ICP algorithm is proposed, which is based on the continuity of the environment around mobile robot. That is, this new algorithm recovers the un-scanned area by estimating feature points in the neighboring two regions based on the continuous environment information. The effectiveness of proposed algorithm has been demonstrated through real experiments of the mobile robot navigation.

Bootstrapping bilinear models of Simple Vehicles

Learning and adaptivity will play a large role in robotics in the future. Two questions are open: (1) in principle, how much it is possible to learn; and (2) in practice, how much should an agent be able to learn. The bootstrapping scenario describes the extreme case in which agents need to learn “everything” from scratch, including a torque-to-pixels model for its robotic body. This paper considers the bootstrapping problem for a subset of the set of all robots. The Simple Vehicles are an idealization of mobile robots equipped with a set of “canonical” exteroceptive sensors: the camera, the range finder and the field sampler. The sensorimotor dynamics of these sensors are derived and shown to be surprising similar. These sensorimotor dynamics are well approximated by a class of nonlinear systems that assume an instantaneous bilinear relation among observations, commands, and changes in the observations. The bilinear approximation is sufficient to guarantee success in the task of generalized “servoing”: driving the observations to a given goal snapshot. Simulations and experiments substantiate the theoretical results. This is the first instance of a bootstrapping agent that can learn the model of the dynamics of a relatively large universe of systems and use the models to solve well-defined tasks, with no parameter tuning or hand-designed features.

Extrinsic calibration of 2D laser rangefinders from perpendicular plane observations

Many applications in the fields of mobile robotics and autonomous vehicles employ two or more 2D laser rangefinders (LRFs) for different purposes: navigation, obstacle detection, 3D mapping or simultaneous localization and mapping. The extrinsic calibration between such sensors (i.e. finding their relative poses) is required to exploit effectively all of the sensor measurements and to perform data fusion. In the literature, most works employing several LRFs obtain their extrinsic calibration from manual measurements or from ad-hoc solutions. In this paper we present a new method to obtain such calibration easily and robustly by scanning perpendicular planes (typically corners encountered in structured scenes), from which geometric constraints are inferred. This technique can be applied to a rig with any number of LRFs in almost any geometric configuration (a minimum of two LRFs whose scanning planes are not parallel is required). Experimental results are presented with synthetic and real data to validate our proposal. A C++ implementation of this method and a dataset are also provided.

Mutual localization in multi-robot systems using anonymous relative measurements

We propose a decentralized method to perform mutual localization in multi-robot systems using anonymous relative measurements, i.e. measurements that do not include the identity of the measured robot. This is a challenging and practically relevant operating scenario that has received little attention in the literature. Our mutual localization algorithm includes two main components: a probabilistic multiple registration stage, which provides all data associations that are consistent with the relative robot measurements and the current belief, and a dynamic filtering stage, which incorporates odometric data into the estimation process. The design of the proposed method proceeds from a detailed formal analysis of the implications of anonymity on the mutual localization problem. Experimental results on a team of differential-drive robots illustrate the effectiveness of the approach, and in particular its robustness against false positives and negatives that may affect the robot measurement process. We also provide an experimental comparison that shows how the proposed method outperforms more classical approaches that may be designed building on existing techniques. The source code of the proposed method is available within the MLAM ROS stack.