This paper presents a control architecture readily suitable for surface treatment tasks such as polishing, grinding, finishing or deburring as carried out by a human operator, with the added benefit of accuracy, recurrence and physical strength as administered by a robotic manipulator partner. The shared strategy effectively couples the human operator propioceptive abilities and fine skills through his interactions with the autonomous physical agent. The novel proposed control scheme is based on task prioritization and a non-conventional sliding mode control, which is considered to benefit from its inherent robustness and low computational cost. The system relies on two force sensors, one located between the last link of the robot and the surface treatment tool, and the other located in some place of the robot end-effector: the former is used to suitably accomplish the conditioning task, while the latter is used by the operator to manually guide the robotic tool. When the operator chooses to cease guiding the tool, the robot motion safely switches back to an automatic reference tracking. The paper presents the theories for the novel collaborative controller, whilst its effectiveness for robotic surface treatment is substantiated by experimental results using a redundant 7R manipulator and a mock-up conditioning tool.

Accidents in which buses or coaches are involved cause thousands of injuries and fatalities every year. To reduce their number and severity, this paper describes an advanced driver assistance system (ADAS) based on a haptic throttle pedal and emergency braking. It also proposes a computationally efficient algorithm with a methodology based on three main concepts: a simplified but accurate vehicle model, an efficient collision detection system considering the driver's intention and pedestrians wandering around the vehicle, and a risk evaluation system to generate warnings and emergency braking signals. Finally, the performance of the proposed ADAS is validated using a driving simulation cabin with a very realistic urban scenario and original elements from real buses. The results show a statistically significant improvement in safety, as the number of collisions and high-risk situations are clearly minimized, the reaction time to press the brake pedal is improved, and the time to collision increased in emergency situations. Implementation of the proposed ADAS into city buses would potentially improve safety, reducing the frequency and severity of accidents in pedestrians.

In this paper, a general formulation of a predictive and multirate (MR) reactive planning method for intelligent vehicles (IVs) is introduced. The method handles path planning and trajectory planning for IVs in dynamic environments with uncertainty, in which the kinodynamic vehicle constraints are also taken into account. It is based on the potential field projection method (PFP), which combines the classical potential field (PF) method with the MR Kalman filter estimation. PFP takes into account the future object trajectories and their associated uncertainties, which makes it different from other look-ahead approaches. Here, a new PF is included in the Lagrange-Euler formulation in a natural way, accounting for the vehicle dynamics. The resulting accelerations are translated into control inputs that are considered in the estimation process. This leads to the generation of a local trajectory in real time (RT) that fully meets the constraints imposed by the kinematic and dynamic models of the IV. The properties of the method are demonstrated by simulation with MATLAB and C++ applications. Very good performance and execution times are achieved, even in challenging situations. In a scenario with 100 obstacles, a local trajectory is obtained in less than 1 s, which is suitable for RT applications.

In this paper, Image-Based Visual Servoing (IBVS) is addressed via a new control structure where reference features are previously filtered based on an estimation of these reference features using standard filters such as Kalman Filter (KF) or a combination of a Kalman Filter and a Smoother (KFS). In this sense, one of the key aspects of the paper is to predict feasible reference features for the low-level IBVS controller. Along the paper, we discuss and analyze the improvements introduced with the new control structure in terms of convergence time and reachability, that is, the ability to converge in complex scenarios such as rotations around the camera axis and large displacements. Validations are also provided through real experimentation with an industrial robot of 6 DOF in eye-in-hand configuration.

This paper describes the design and implementation of a novel inspection system for detecting defects on car bodies based on artificial vision, implemented in Ford Factory at Almussafes (Spain). The system is based on the principle of performing a lightning sweeping with static imagining system, which causes shadows surrounding defects when merging consecutive images, coined as defect augmentation phenomena. As a result, we can detect millimetric defects of 0.3mm diameter or greater with different shapes which were very hard to detect with existing technology without that phenomena. The project has generated two PTC patents, the first one protects the defect augmentation phenomena, while the second protects the industrial system itself. The main innovation of this industrial project is the development of a system that improves in almost 100% the human inspection. As a consequences, it reduces the number of invalid vehicles, energy consumption, saving painting which also implies a significant cost reduction. It also improves working conditions for workers by reducing ocular fatigues.

The main contribution of the paper is to provide a method to generate continuous curvature paths in order to converge to a line, based on combinations of clothoids with line segments and circular arcs. Different kinds of continuous curvature paths have been defined in order to solve problems with different complexity. The type of paths that we have generated get the benefits of higher comfort and safety. Generated paths take into account lower and upper bounds of sharpness and curvature simultaneously, while it is not the case of other continuous curvature paths such as Elementary and BiElementary paths. Wheeled mobile robots following a path with continuous curvature may also get benefit on wheels slippage reduction and low odometry errors, since transitions are softer with constant curvature rates.

The main contribution of the paper is to demonstrate the benefits of a new kinematic control method within the Pure-Pursuit framework, based on continuous-curvature paths, against a classic approach to Pure-Pursuit methods commonly used in mobile robotics. The new method can be applied in a generic path following problem and it is based on combinations of clothoids with line segments and circular arcs, providing a general curvature profile. The main advantage is that generated paths take into account lower and upper bounds of curvature and sharpness which implies smoother and safer trajectories, not considered in other methods. It is shown that the proposed method has better performance with respect to settling time, overshoot and little influence on design parameters. As application, a vision-based path following continuous-curvature control problem has been implemented on an industrial forklift vehicle in order to show the advantage of the method.

The paper addresses the problem of evaluating AGVs with different degrees of autonomy by defining benchmark tools to grade the performance of each approach. Based on the proposed benchmark, different experiments have been performed from manual driving to autonomous navigation, at different velocities and a scenario containing wide and narrow corridors, small and large isles, rooms, slalom-like parts requiring zig-zag maneuvering and static objects. However, this benchmark is also applicable to dynamic environments, including moving objects and other vehicles. In particular, experiments have been used for evaluating the performance of AGVs, in terms of robustness, efficiency, safety and comfortability. The underlying objective is to evaluate the potential advantages of manual-assisted driving as well as autonomous navigation against standard manual driving. To obtain valid and significant results, more than 180 experiments have been carried out on each approach.

This paper presents some intermediate results of a research project regarding with auto-guided vehicles (AGVs) and vision-based guidance. The paper describes three different degrees of autonomy, form the basic one to the full autonomous one. In particular, the paper focuses on the intermediate degree of autonomy, named as ¿guided¿ driving, which includes several types of application under the same approach such as manual-assisted driving, teleoperated driving and a vision-based line tracking application. The paper describes on the line-tracking problem with AGVs where a simple and robust line detection algorithm has been described and implemented on an embedded vision system. It also describes a line tracking control algorithm, which has been validated through experimental and simulation data. Finally, the paper discusses future work in the context of the project.

This paper presents a novel technique for implementing Clothoidal real-time paths for mobile robots. As first step, rational Bezier curves are obtained as approximation of the Fresnel integrals. By rescaling, rotating and translating the previously computed RBC, an on-line Clothoidal path is obtained. In this process, coefficients, weights and control points are kept invariant. This on-line approach guarantees that an RBC has the same behavior as the original Clothoid using a low curve order. The resulting Clothoidal path allows any two arbitrary poses to be joined in a plane. RBCs working as Clothoids are also used to search for the shortest bounded-curvature path with a significant computational cost reduction. In addition to this, the proposed technique is tested on a real mobile robot for trajectory generation and kinematic control. To the authors' knowledge, the present approach is the first technique which allows real-time Clothoidal path computation.

In this paper, a multi-rate predictive artificial potential field method for dynamic and uncertain environments is described. It is based on the combination of classical Artificial Potential Field methods (APF) with Multi-rate Kalman Filter estimations (MKF), which takes into account present and future obstacle locations within a temporal horizon. By doing that, position uncertainty of obstacles is considered in the avoidance algorithm. This implies anticipation to the movement of the obstacles and its consideration in the path planning strategy. In this paper, forces derived from the potential field are taken as control inputs for the system model as well as considered in the Kalman Filter estimation. This leads to the generation of a local trajectory that fully meets the restrictions imposed by the kinematic model of the robot.

This paper presents a set of robust and efficient algorithms with O(N) cost for the following situations: 1) object detection with a laser ranger; 2) mobile robot pose estimation and 3) an improved Monte-Carlo localization (MCL) method using multi-rate techniques. Object detection is mainly based on a novel multiple line fitting method for wall detection with regular constrained angles. Columns are also detected based on conventional circle fitting. In addition, two mobile robot self-localization methods are based on detected walls and columns respectively. These methods perform an efficient estimation based on LS sense and they can provide global pose estimation under assumption of known data-association. Moreover, the standard MCL method has been extended by considering the asynchronous sampling of sensors and actuators. Experimental results show that significant improvements in pose estimation can be obtained, since the algorithm runs at the fastest possible sampling frequencies for each process. This improvement allows the same accuracy as the standard MCL with fewer particles and lower computational cost.

In this paper, we present a set of robust and efficient algorithms with O(N) cost for the following situations: object detection with a laser ranger; mobile robot pose estimation and a FastSLAM improved implementation. Objected detection is mainly based on a novel multiple line fitting method, related with walls at the environment. This method assumes that walls at the environment constitute a regular constrained angles. A line-based pose estimation method is also proposed, based on least-squares (LS). This method performs the matching of detected lines and estimated map lines and it can provide the global pose estimation under assumption of known data-association. FastSLAM 1.0 has been improved by considering the estimated pose with the LS-approach to re-allocate each particle of the posterior distribution. This approach has a lower computational cost than EKF approach in FastSLAM 2.0. The three algorithms have been combined in order to perform an efficient self-localization and map building process, tested for indoor environments with real data. And results show that the ideas proposed in this paper could aim in closing the loop and also to improve the overall estimation performance

In this paper, we present a set of robust and efficient algorithms with O(N) cost for object detection and pose estimation with a laser ranger based on regular geometrical maps. Firstly, a multiple line fitting method is described, related with walls at the environment, which minimizes the sum of squared distances for contiguous lines and constitutes a global pattern with regular constrained angles. Secondly, beacons, related with columns at the environment, are estimated with the circle algebraic method. Two pose estimation methods are presented, based on detected beacons and lines. These methods have proved to be robust and faster than other methods. In addition to this, a redundant kinematic model is used for dead-reckoning sensors, which may improve estimations. Finally, the robot localization is performed with an EKF during normal navigation and a global localization method, based on Monte-Carlo simulation methods, at initialization.

This work presents a multi-rate fusion model, which exploits the complimentary properties of visual and inertial sensors for egomotion estimation in applications such as robot navigation and augmented reality. The sampling of these two sensors is described with size-varying input and output equations without assumed synchronicity and periodicity of measurements. Data fusion is performed with two different multi-rate (MR) filter models, an extended (EKF) and an unscented Kalman filter (UKF). A complete dynamic model for the 6D-tracking task is given together with a method to calculate the dependencies of the covariance matrices. It is further shown that a centripetal acceleration model and the precise description of quaternion prediction for a constant velocity model highly improve the estimation error for rotary motions. The comparison demonstrates that the MR-UKF provides better estimation results at higher computational costs.

In this paper, extended Kalman and unscented Kalman filters are developed for multi-rate systems in the context of the SLAM problem. These multi-rate filters have been extensively compared and tested with experimental data taken from a parking lot. Both multi-rate filters have improved the estimation with respect to the conventional single-rate Kalman filter. A performance index is introduced, showing that EKF gives better performance than UKF. In order to complete the SLAM solution, well-known techniques for feature extraction, data association and map building have been also implemented.

This paper presents a global solution developed for integrating the warehouses transportation system based on a fleet of automated industrial forklifts in the factory management. Each industrial forklift has been completely automated using external and internal sensors. An ERP system, as information system, has been installed in order to handle the whole factory management process. A SCADA application has also been implemented for monitoring the industrial vehicles in the plant. It also acts as an interface between the ERP and the process itself. Finally, three vehicle transportation modes have been developed and validated in the system: teleoperation, path-tracking and line tracking.

In this paper, a new method for generating a set of collision-free maneuvers for unmanned vehicles in automated highway systems is introduced. The low computational cost of the proposed maneuver planner allows its execution as frequent as the new positions and speeds of vehicles (obstacles) in sight are received. This planner is based on the computation of the minimum translational distance between two mobile objects. This distance is then used for predicting and avoiding a collision, generating (if it is possible and safe), overtaking (changing to the left and right lane), braking and keeping-the-lane maneuvers. These maneuvers can be provided to a decision-making process for selecting the most suitable one.

A new technique for collision detection based on the application of the Hough transform is presented in this paper. As a consequence of applying this technique, a method for computing the minimum translational distance (MTD) between two spherically extended polytopes is introduced. This distance algorithm is based on the Gilbert-Johnson-Keerthi (GJK) algorithm. Parameters returned by the MTD algorithm hold collision-free configurations that avoid a predicted collision.

A methodology for computing a collision-free trajectory for mobile robots amidst moving objects is presented. This planner is based on a technique for computing the minimum translational distance between two mobile objects. This distance is then used for predicting and avoiding a collision. The computation of this distance is based on the application of the GJK algorithm to a particular subset of the Minkowski difference set of the involved objects. This subset states the separation or penetration distance between two objects along their given motions. When a collision is predicted, a collision-free intermediate temporal-position is generated avoiding such a collision.

Presents an approach to solving the linear quadratic Gaussian problem for controlling MIMO multirate sampled-data systems. The problem is split into three parts: first a generalized discrete multirate model is computed, second the solution of the linear quadratic regulator problem is computed and then a multirate Kalman filter is obtained. The model is composed by a time invariant single rate part, involving the basic process dynamics and a periodic one determined by the external sampling strategy. This allows an easy analysis of the effects of the multirate sampling on the plant. The approach has been applied to solve problems of self-location and path tracking in mobile robots. As is common in many auto-guided vehicles the sensory system is very complex, so some kind of sensor data fusion has been implemented.

A novel and fast technique for solving the basic motion-planning problem is introduced. Obstacle avoidance is based on the computation of minimum translational distances that allows one to obtain collision-free intermediate configurations. When path planning is constrained to two degrees of freedom, Hough transform is applied to such configurations. After that, several collision-free paths are generating by joining different sets of selected intermediate configurations. These selections are obtained by applying the technique called minimum volume locus. Complexity of this path planner is linear with the number of obstacles. Object representation is based on spherically extended polytopes.

An object representation scheme which allows fast Euclidean distance calculation is presented. The object model extends the polytope model by representing objects as the convex hull of a finite set of spheres. Objects with rounded surfaces can be represented with far fewer vertices than would be required with the polytope model. An algorithm for calculating distances between objects is developed which is linear in the total number of spheres specifying the two objects. To determine the efficiency of the distance computation algorithms, a program was written and tested using randomly generated objects with varying numbers of vertices.>