Field and Service Robotics

The first functional load-carrying and energetically autonomous exoskeleton was demonstrated at U.C. Berkeley, walking at the average speed of 1.3 m/s while carrying a 34 kg (75 lb) payload. Four fundamental technologies associated with the Berkeley Lower Extremity Exoskeleton (BLEEX) were tackled during the course of this project. These four core technologies include: the design of the exoskeleton architecture, control schemes, a body local area network (bLAN) to host the control algorithm and an on-board power unit to power the actuators, sensors and the computers. This article gives an overview of one of the control schemes. The analysis here is an extension of the classical definition of the sensitivity function of a system: the ability of a system to reject disturbances or the measure of system robustness. The control algorithm developed here increases the closed loop system sensitivity to its wearer’s forces and torques without any measurement from the wearer (such as force, position, or electromyogram signal). The control method has little robustness to parameter variations and therefore requires a relatively good dynamic model of the system.

This paper describes an algorithm that tracks and localizes a helicopter using a ground-based trinocular camera array. The three cameras are placed independently in an arbitrary arrangement that allows each camera to view the helicopter’s flight volume. The helicopter then flies an unplanned path that allows the cameras to self-survey utilizing an algorithm based on structure from motion and bundle adjustment. This yields the camera’s extrinsic parameters allowing for real-time positioning of the helicopter’s position in a camera array based coordinate frame. In fielded experiments, there is less than a 2

m

RMS tracking error and the update rate of 20

Hz

is comparable to DGPS update rates. This system has successfully been integrated with an IMU to provide a positioning system for autonomous hovering.

Performing reliable localisation and navigation within highly unstructured underwater coral reef environments is a difficult task at the best of times. Typical research and commercial underwater vehicles use expensive acoustic positioning and sonar systems which require significant external infrastructure to operate effectively. This paper is focused on the development of a robust vision-based motion estimation technique using low-cost sensors for performing real-time autonomous and untethered environmental monitoring tasks in the Great Barrier Reef without the use of acoustic positioning. The technique is experimentally shown to provide accurate odometry and terrain profile information suitable for input into the vehicle controller to perform a range of environmental monitoring tasks.

Awareness of pedestrians, other vehicles, and other road obstacles is key to driving safety, and so their detection is a critical need in driver assistance research. We propose using a model-based approach which can either directly segment the disparity to detect obstacles or remove the road regions from an already segmented disparity map. We developed two methods for segmentation: first, by directly segmenting obstacles from the disparity map; and, second by using morphological operations followed by a robust model fitting algorithm to reject road segments after the segmentation process. To test the success of our methods, we have tested and compared them with an available method in the literature.

In this paper, we present a new adaptation of the regular polygon detection algorithm for real-time road sign detection for autonomous vehicles. The method is robust to partial occlusion and fading, and insensitive to lighting conditions. We experimentally demonstrate its application to the detection of various signs, particularly evaluating it on a sequence of roundabout signs taken from the ANU/NICTA vehicle. The algorithm runs faster than 20 frames per second on a standard PC, detecting signs of the size that appears in road scenes, as observed from a camera mounted on the rear-vision mirror. The algorithm uses the symmetric nature of regular polygonal shapes, we also use the constrained appearance of such shapes in the road scene to the car in order to facilitate their fast, robust detection.

Autonomous navigation using natural landmarks in an unexplored environment is a very difficult problem to handle. While there are many techniques capable of matching pre-defined objects correctly, few of them can be used for real-time navigation in an unexplored environment. One important unsolved problem is to efficiently select a minimum set of usable landmarks for localisation purposes. This paper presents a method which minimises the number of landmarks selected based on texture descriptors. This enables localisation based on only a few distinctive landmarks rather than handling hundreds of irrelevant landmarks per image. The distinctness of a landmark is calculated based on the mean and covariance matrix of the feature descriptors of landmarks from an entire history of images. The matrices are calculated in a training process and updated during real-time navigation.

We present a biologically inspired active vision system that incorporates two modes of perception. A peripheral mode provides a broad and coarse perception of where mass is in the scene in the vicinity of the current fixation point, and how that mass is moving. It involves fusion of actively acquired depth data into a 3D occupancy grid. A foveal mode then ensures coordinated stereo fixation upon mass/objects in the scene, and enables extraction of the mass/object using a maximum a-posterior probability zero disparity filter. Foveal processing is limited to the vicinity of the camera optical centres. Results for each mode and both modes operating in parallel are presented. The regime operates at approximately 15

Hz

on a

3GHz

single processor PC.

This paper describes a system for automatic marking of floors. Such systems can be used for example when marking the positions of stands for a trade fair or exhibition. Achieving a high enough accuracy in such an environment, characterized by very large open spaces, is a major challenge. Environmental features will be much further away then in most other indoor applications and even many outdoor applications.

A SICK LMS 291 laser scanner is used for localization purposes. Experiments show that many of the problems that are typically associated with the large beam width of ultra sonic sensors in normal indoor environments manifest themselves here for the laser because of the long range.

The system that is presented has been in operation for almost two years to date and has been used for every exhibition in the three main exhibition halls at the Stockholm International Fair since then. The system has speeded up the marking process significantly. For example, what used to be a job for two men over eight hours now takes one robot monitored by one man four hours to complete.

This paper develops a theory for the decentralised estimation of angular profiles for point characterisation. Angular profiles are an example of spatially distributed characterisation which may be further processed for classification and identification. This paper also describes the development of a vision system for integration into a decentralised data fusion system for tracking and characterisation. Visual results are presented showing the field setup and sensor observations.

The need for reliable maps of subterranean spaces too hazardous for humans to occupy has motivated the development of robotic mapping tools. For such systems to be fully autonomous, they must be able to deal with all varieties of subterranean environments, including those containing loops. This paper presents an approach for an autonomous mobile robot to determine if the area currently being explored has been previously visited. Combined with other techniques in topological mapping, this approach will allow for the fully autonomous general exploration of subterranean spaces. Data collected from a research coal mine is used to experimentally verify our approach.

For underground mining operations human operated LHD vehicles are typically used for transporting ore. Because of security issues and of the cost of human operators, alternative solutions such as tele-operated vehicles are often in use. Tele-operation, however, leads to reduced efficiency, and it is not an ideal solution. Full automation of the LHD vehicles is a challenging task, which is expected to result in increased operational efficiency, cost efficiency, and safety. In this paper, we present our approach to a fully automated solution currently under development. We use a fuzzy behavior-based approach for navigation, and develop a cheap and robust localization technique based on the deployment of inexpensive passive radio frequency identification (RFID) tags at key points in the mine.

In this paper an existing method for indoor Simultaneous Localisation and Mapping (SLAM) is extended to operate in large outdoor environments using an omnidirectional camera as its principal external sensor. The method, RatSLAM, is based upon computational models of the area in the rat brain that maintains the rodent’s idea of its position in the world. The system uses the visual appearance of different locations to build hybrid spatial-topological maps of places it has experienced that facilitate relocalisation and path planning. A large dataset was acquired from a dynamic campus environment and used to verify the system’s ability to construct representations of the world and simultaneously use these representations to maintain localisation.

D-SLAM algorithm first described in [

1.

] allows SLAM to be decoupled into solving a non-linear static estimation problem for mapping and a three-dimensional estimation problem for localization. This paper presents a new version of the D-SLAM algorithm that uses an absolute map instead of a relative map as presented in [

1.

]. One of the significant advantages of D-SLAM algorithm is its

O

(

N

) computational cost where

N

is the total number of features (landmarks). The theoretical foundations of D-SLAM together with implementation issues including data association, state recovery, and computational complexity are addressed in detail. Evaluation of the D-SLAM algorithm is provided using both real experimental data and simulations.

This paper presents a new generalisation of simultaneous localisation and mapping (SLAM). SLAM implementations based on

extended Kalman filter

(EKF) data fusion have traditionally relied on simple geometric models for defining landmarks. This limits EKF-SLAM to environments suited to such models and tends to discard much potentially useful data. The approach presented in this paper is a marriage of EKF-SLAM with scan correlation. Instead of geometric models, landmarks are defined by templates composed of raw sensed data, and scan correlation is shown to produce landmark observations compatible with the standard EKF-SLAM framework. The resulting

Scan

-SLAM combines the general applicability of scan correlation with the established advantages of an EKF implementation: recursive data fusion that produces a convergent map of landmarks and maintains an estimate of uncertainties and correlations. Experimental results are presented which validate the algorithm.

We present a probabilistic technique for alignment and subsequent change detection using range sensor data. The alignment method is derived from a novel, non-rigid approach to register point clouds induced by pose-related range observations that are particularly erroneous. It allows for high scan estimation errors to be compensated distinctly, whilst considering temporally successive measurements to be correlated. Based on the alignment, changes between data sets are detected using a probabilistic approach that is capable of differentiating between likely and unlikely changes. When applied to observations containing even small differences, it reliably identifies intentionally introduced modifications.

Elevation maps are a popular data structure for representing the environment of a mobile robot operating outdoors or on not-flat surfaces. Elevation maps store in each cell of a discrete grid the height of the surface the corresponding place in the environment. The use of this

$$ 2\tfrac{1} {2}$$

-dimensional representation, however, is disadvantageous when it is used for mapping with mobile robots operating on the ground, since vertical or overhanging objects cannot be represented appropriately. Such objects furthermore can lead to registration errors when two elevation maps have to be matched. In this paper we propose an approach that allows a mobile robot to deal with vertical and overhanging objects in elevation maps. We classify the points in the environment according to whether they correspond to such objects or not. We also describe a variant of the ICP algorithm that utilizes the classification of cells during the data association. Experiments carried out with a real robot in an outdoor environment demonstrate that the scan matching process becomes significantly more reliable and accurate when our classification is used.

In this paper, a method of obtaining vehicle hypothesis based on laser scan data only is proposed. This is implemented on the robotic vehicle, CyCab, for navigation and mapping of the static car park environment. Laser scanner data is used to obtain hypothesis on position and orientation of vehicles with Bayesian Programming. Using the hypothesized vehicle poses as landmarks, CyCab performs Simultaneous Localization And Mapping (SLAM). A final map consisting of the vehicle positions in the car park is obtained.

This paper introduces the structure of wavelet occupancy grids (WavOGs) as a tool for storing occupancy grids in a compact way. We have shown that WavOGs provide a continuous semantics of occupancy through scaled spaces. In accordance with the theoretical properties of wavelets, our experiments have validated that WavOGs allow major memory gains. WavOG as a compact multi-scaled tool provides an efficient framework for the various algorithms that use OGs such as robot navigation, spatio-temporal classification or multiple target-tracking. In future works we plan to apply WavOGs to the monitoring of urban traffic over large areas.

In this paper, we present recent results with using range from radio for mobile robot localization. In previous work we have shown how range readings from radio tags placed in the environment can be used to localize a robot. We have extended previous work to consider robustness. Specifically, we are interested in the case where range readings are very noisy and available intermittently. Also, we consider the case where the location of the radio tags is not known at all ahead of time and must be solved for simultaneously along with the position of the moving robot. We present results from a mobile robot that is equipped with GPS for ground truth, operating over several km.

In this paper we describe experiments with networks of robots and sensors in support of search and rescue and first response operations. The system we consider includes a network of Mica Mote sensors that can monitor temperature, light, and the movement of the structure on which they rest. We also consider an extension to chemical sensing in simulation only. An ATRV-Mini robot is extended with a Mote sensor and a protocol that allows it to interact with the network. We present algorithms and experiments for aggregating global maps in sensor space and using these maps for navigation. The sensor experiments were performed outdoors as part of a Search and Rescue exercise with practitioners in the field.

This paper presents a data-fusion and interpretation system for operation of an Autonomous Ground Vehicle (AGV) in outdoor environments. It is a practical implementation of a new model for machine perception and reasoning, which has its true utility in its applicability to increasingly unstructured environments. This model provides a cohesive, sensor-centric and probabilistic summary of the available sensory data and uses this richly descriptive data to enable robust interpretation of a scene. A general model is described and the development of a specific instance of it is described in detail. Preliminary results demonstrate the utility of the approach in very large, unstructured, outdoor environments.

We propose a principled method to create a search space for constrained motion planning, which efficiently encodes only feasible motion plans. The space of possible paths is encoded implicitly in the connections between states, but only feasible and only local connections are allowed. Furthermore, we propose a systematic method to generate a near-minimal set of spatially distinct motion alternatives. This set of motion primitives preserves the connectivity of the representation while eliminating redundancy — leading to a very efficient structure for motion planning at the chosen resolution.

This paper discusses some issues for generating points of contact on object grasping by multifingered robot hands. To address these issues, we present a general algorithm based on computer vision techniques for determining grasping points through a sequence of processes: (1) object’s visual features, we apply some algorithms for extracting vertices, edges, object’s contours, (3) modeling the point of contact by a bounded polytope, (3) based on these features, the developed algorithm starts by analysing the object’s contour to generate a set of contact points that guarantee the force-closure grasps condition. Finally, we briefly describe some experiments on a humanoid robot with a stereo camera head and an anthropomorphic robot hand within the “Center of Excellence on Humanoid Robots: Learning and co-operating Systems” at the University of Karlsruhe and the Forschungszentrum Karlsruhe.

A key aspect of service robotics for everyday use is the motion in close proximity to humans. It is essential that the robot exhibits a behavior that signals safety of motion and awareness of the persons in the environment. To achieve this, there is a need to define control strategies that are perceived as socially acceptable by users that are not familiar with robots. In this paper a system for navigation in a hallway is presented, in which the rules of proxemics are used to define the interaction strategies. The experimental results show the contribution to the establishment of effective spatial interaction patterns between the robot and a person.

Motion prediction for objects which are able to decide their trajectory on the basis of a planning or decision process (e.g. humans and robots) is a challenging problem. Most existing approaches operate in two stages: a) learning, which consists in observing the environment in order to identify and model possible motion patterns or plans and b) prediction, which uses the learned plans in order to predict future motions. In existing techniques, learning is performed off-line, hence, it is impossible to refine the existing knowledge on the basis of the new observations obtained during the prediction phase. This paper proposes a novel learning approach which represents plans as Hidden Markov Models and is able to estimate the parameters and structure of those models in an incremental fashion by using the Growing Neural Gas algorithm. Our experiments demonstrate that the technique works in real-time, is able to operate concurrently with prediction and that the resulting model produces long-term predictions.

This paper presents a novel design and a motion planner for a semipassive mobile robot. The robot consists of an upper circular body and three identical semi-passive driving mechanisms. Each mechanism consists of a passive wheel that can freely roll, a rotation actuator along the normal axes and a linear actuator for motion along the radial direction of the upper body center. The robot is equipped with an inclinometer to measure the surface slope. Each wheel is also equipped with a rotational encoder to measure roll. Using an odometric model, data from these encoders determines vehicle position. Kinematic analysis provides tools for designing a motion path that steers the robot to the desired location, and determines the singular configurations. Due to the passive roll, there is no longitudinal slippage, and lateral slippage is determined from the kinematic and odometric models. This enables accurate and reliable localization even with slippage. A gait pattern planer for downhill, as well as horizontal and uphill surfaces is presented. A prototype robot has been built and field tested. Experimental results verify the suggested models.

In our current research, we are developing a holonomic mobile vehicle which is capable of running over the step. This system realizes omni-directional motion on flat floor using special wheels and passes over the step in forward or backward direction using the passive suspension mechanism. This paper proposes a new wheel control method of the vehicle according to its body configuration for passing over the step. The developed vehicle utilizes the passive suspension mechanism connected by two free joints that provide to change the body configuration on the terrain condition. Therefore, it is required to coordinate the suitable rotation velocity of each wheel according to its body configuration. In our previous work, the vehicle motion during step-climbing was discussed and moving velocity of each wheel was derived. In this paper, we adapt these results to wheel control and derived rotation velocity reference of each wheel. The performance of our proposed method is verified by the computer simulations and experiments using our prototype vehicle.

The design and control of high power water hydraulic robotic systems presents unique challenges. This paper describes some of those challenges and the techniques used to overcome them to develop a unique system for an autonomous robotic drilling machine for underground mining.

In most search and rescue or reconnaissance missions involving field robots the requirements of the operator being mobile and alert to sudden changes in the near environment, are just as important as the ability to control the robot proficiently. This implies that the GUI platform should be light-weight and portable, and that the GUI itself is carefully designed for the task at hand. In this paper different platform solutions and design of a user-friendly GUI for a packbot will be discussed. Our current wearable system will be presented along with some results from initial field tests in urban search and rescue facilities.

Mobile robots used in the human-robot coexisting environment are required to perform continuous works without human maintenance. On the other hand, they need a rechargeable batteries that require charge, generally. Therefore, an autonomous battery-charging for mobile robots has a big advantage for performing continuous works. However, installation of exclusive use of battery-charging-stations requires much cost.

To improve this situation, we aim to realize an autonomous battery-charging motion for a mobile manipulator using conventional electrical outlets. In this motion, the robot is navigated to a front of an outlet, and a plug attached at the tip of the manipulator is controlled to insert into the outlet. To realize the motion, we implemented “distance transform method” for navigation, and a motion of plug insertion using force feedback control.

In this paper, we explain the above implementations, and we discuss advantages and limitations of such motions.

In this paper we propose an innovative robot remote control method, a synthesized scene recollection method, which provides the operator with a bird’seye view image of the robot in an environment which is generated by using position and orientation information of the robot, stored image history data captured by a camera mounted on the robot, and the model of the robot. This method helps the operator to easily recognize the situation of the robot even in unknown surroundings and enables the remote operation ability of a robot to be improved. This method is mainly based on two technologies, robot positioning and image synthesis. In this paper we use scan matching of laser rangefinder’s scan data for robot positioning and realized self-contained implementation of the proposed method in 2D horizontal plane.

In this paper, a newly initiated project of networked robotic system for disaster mitigation is introduced. In this project, multiple robots are coordinately operated through ad-hoc wireless communication network, including satellite-based IP communication link, for surveillance tasks at a disaster site. The robot system consists of a large-scale outdoor robot to serve as a carrier of small robots and a fleet of small robots to be deployed at a specific spot such as an inside of a building complex. A combination of a laser range scanner and an omni-directional camera is used to acquire high resolution 3D geometry data and rendering images. Those data and images are displayed using

Mixed Reality

(MR) technology at a remote site to provide an overall picture for operation managers with high fidelity. This paper presents our initial experiments using a robot test bed with an emphasis on remote operation over rough terrain and for acquisition of high resolution 3D geometry data and telepresence using MR technology.

This paper presents a practical method for small-scale VTOL

3

design. It helps for elements selection and dimensioning. We apply the latter to design a fully autonomous quadrotor with numerous innovations in design methodology, steering and propulsion achieving 100% thrust margin for 30 min autonomy. The robot is capable of rotational and translational motion estimation. Finally, we derive a nonlinear dynamics simulation model, perform a simulation with a PD test controller and test successfully the robot in a real flight. We are confident that “OS4” is a significant progress towards intelligent miniature quadrotors.

The Autonomous Systems Lab of EPFL

3

is developing, within the framework of an ESA program, an ultra-lightweight solar autonomous model airplane called Sky-Sailor with embedded navigation and control systems. The main goal of this project is to jointly undertake research on navigation, control of the plane and also work on the design of the structure, the energy generation system. The airplane will be capable of continuous flight over days and nights, which makes it suitable for a wide range of applications.

This paper details the control and guidance architecture for the T-Wing tail-sitter unmanned air vehicle, (UAV). The vehicle uses a mixture of classical and LQR controllers for its numerous low-level and guidance control loops. Different controllers are used for the vertical, horizontal and transition flight modes, glued together with supervisory mode-switching logic. This allows the vehicle to achieve autonomous waypoint navigation throughout its flight-envelope. The control design for the T-Wing is complicated by the large differences in vehicle dynamics between vertical and horizontal flight; the difficulty of accurately predicting the low-speed vehicle aerodynamics; and the basic instability of the vertical flight mode. This paper considers the control design problem for the T-Wing in light of these factors. In particular it focuses on the integration of all the different types and levels of controllers in a full flight-vehicle control system.

This paper presents a modular architecture for the implementation and evaluation of control strategies for a team of autonomous Unmanned Aerial Vehicles (UAVs). The architecture is designed to be a test bed for control algorithms ranging from low level vehicle control to higher level multi-vehicle mission control. The architecture is being implemented on a team of Brumby Mk III fixed wing UAVs. Particular focus is on the testing procedure, through simulation, HardWare- In-the-Loop testing and real time flight tests. Results from recent flight testing of a path-following guidance algorithm used for feature orbiting are shown and future implementation of cooperative UAV strategies are presented.

Trajectory generation has traditionally been formulated on the assumption that the environment is flat. On rough terrain, however, deviations of the angular velocity vector from the vertical lead to errors which accumulate in a manner similar to the accumulation of attitude errors in odometry. In practice, feedback control can compensate for these errors in modeling by adjusting the path in real time. In many realistic cases, however, the 3D shape of the terrain is known beforehand, so it is possible to incorporate terrain shape into the predictive model — rather than treat it as an unknown disturbance. This paper presents an algorithm for trajectory generation which compensates for terrain shape in a predictive fashion. The numerical implementation makes it adaptable readily to a broad class of vehicles and even a broad class of predictable disturbances beyond terrain shape. In support of the latter, we demonstrate the ability to invert models of actuator dynamics and wheel slip concurrently with 3D terrain. An example application for a the Rocky 7 Mars rover platform is presented.

This paper presents an outdoor mobile robot capable of high-speed navigation in outdoor environments. Here we consider the problem of a robot that has to follow a designated path at high speeds over undulating terrain. It must also be perceptive and agile enough to avoid small obstacles. Collision avoidance is a key problem and it is necessary to use sensing modalities that are able to operate robustly in a wide variety of conditions. We report on the sensing and control necessary for this application and the results obtained to date.

Foot groping is one way to evaluate the stability of footholds for legged locomotives on rough terrain. For further acquisition of ground information, we installed active ankles with two active joints on the experimental quadruped vehicle, RoQ2. To compensate the loss of passive adaptation of ankles to terrain, active adaptation using COF estimation is implemented. COF is a center of pressure on a sole and estimated by sole sensor, which consists of four FSRs. Sole sensors for COF can determine the sole plane when adapting to rough terrain. This paper also shows that our new proposition can detect an edge of a beam or a step on the ground without thrusting a foot to the objects.

A technique for identifying lumped soil parameters on-line while traversing with a tracked unmanned ground vehicle (UGV) on an unknown terrain is presented. This paper shows the multi-solution problem when identification of soil parameters — cohesion (

c

), shear deformation modulus (

φ

), and shear deformation modulus (

K

) are to be attempted using the track-terrain interaction dynamics model. The initiation of the idea of lumping the cohesion and internal friction angle terms and treating them as a single parameter to solve this problem is presented. The technique used for lumped soil parameter identification is based on the Newton Raphson method. This method is proved to be very effective in terms of prediction accuracy, computational speed, and robustness to initial conditions and noise. These identified lumped soil parameters can be used to increase the autonomy of a tracked UGV. The technique presented in this paper is general and can be applied to any tracked UGV.

The intent of this paper is to show how the accuracy of 3D position tracking can be improved by considering rover locomotion in rough terrain as a holistic problem. An appropriate locomotion concept endowed with a controller minimizing slip improves the climbing performance, the accuracy of odometry and the signal/noise ratio of the onboard sensors. Sensor fusion involving an inertial measurement unit, 3D-Odometry, and visual motion estimation is presented. The experimental results show clearly how each sensor contributes to increase the accuracy of the 3D pose estimation in rough terrain.

In the near future, off-road mobile robots will feature high levels of autonomy which will render them useful for a variety of tasks on Earth and other planets. Many terrestrial applications have a special demand for robots to possess similar qualities to human-driven machines: high speed and maneuverability. Meeting these requirements in the design of autonomous robots is a very hard problem, partially due to the difficulty of characterizing the natural terrain that the vehicle will encounter and estimating the effect of these interactions on the vehicle. Here we present a dynamic traction model that describes vehicle braking on a variety of terrestrial soil types and in a wide range of natural landscapes and vehicle velocities. This model was developed empirically, it is simple yet accurate and can be readily used to improve model-predictive planning and control. The model encapsulates the specifics of wheel-terrain interaction, offers a good compromise between accuracy and real-time computational efficiency, and allows straight-forward consideration of vehicle dynamics.

This paper describes automation of the digging cycle of a mining rope shovel which considers autonomous dipper (bucket) filling and determining methods to detect when to disengage the dipper from the bank. Novel techniques to overcome dipper stall and the online estimation of dipper “fullness” are described with in-field experimental results of laser DTM generation, machine automation and digging using a 1/7th scale model rope shovel presented.

The Emschergenossenschaft based in Germany is currently planning the Emscher sewer system, arguably the largest residential water management project in Europe in years to come. In 2002, the Emschergenossenschaft engaged the Fraunhofer Institute for Factory Operation and Automation (IFF) in Magdeburg, Germany as the general contractor to develop automatic inspection and cleaning systems to meet the requirements stipulated by the legal guidelines. The systems must operate continuously in a sewer line, which has diameters ranging from 1400 to 2800 mm and is partially filled, 25% at minimum, all the time. To construct the Emscher sewer system, the Emschergenossenschaft favors a one-pipe line in long sections. A walk-through or inspection by personnel is impossible in every phase. The Fraunhofer Institute IFF has completed an extensive concept study for the inspection and cleaning systems and has developed as prototypes all systems for motion along the sewer and all sensor systems, achieving a new quality of inspection under these difficult conditions. This article describes significant project results and important components of the inspection and cleaning systems such as the inspection systems, pipe axis measurement, system positioning and sensor systems for damage detection.

Fundamental for the development of the inspection systems are a detailed inspection going far beyond the video inspection common today and the capability of taking comparative measurements throughout the sewer system’s period of operation to describe the development of damage.

The objective of this research is the replacement of hand weeding in organic farming by a device working autonomously at field level. The autonomous weeding robot was designed using a structured design approach, giving a good overview of the total design. A vehicle was developed with a diesel engine, hydraulic transmission, four-wheel drive and four-wheel steering. The available power and the stability of the vehicle does not limit the freedom of research regarding solutions for intra-row weed detection and weeding actuators. To fulfill the function of navigation along the row a new machine vision algorithm was developed. A test in sugar beet in a greenhouse showed that the algorithm was able to find the crop row with an average error of less than 25 mm. The vehicle is a versatile design for an autonomous weeding robot in a research context. The result of the design has good potential for autonomous weeding in the near future.

In this paper, as a part of research work on the autonomous loading operation by wheel loader at surface mines or construction working places, a method of path generation for wheel loader will be described. V shape path connecting between the scooping position and the loading position consists of clothoid curves and straight lines. Each length of line segments are optimized in path generation procedure. The scooping direction is determined based on the estimation of resistance force applied on the bucket during scooping motion, by using simplified shape model of pile and bucket trajectory model. Proposed method is installed on the experimental model. Shape of the pile is measured by a stereo-vision system. For giving scooping position, scooping direction giving the least moment on the bucket is selected. By this method, appropriate path is generated.

In many respects traditional automation in the forest-machine industry has reached an upper limit, since the driver already has to deal with an excess of information and take too many decisions at a very high pace. To further automation still, introduction of semi-autonomous and autonomous functions are expected and considered necessary. This paper describes an ongoing project along these ideas. We describe the development of the hardware and software of an unmanned shuttle that shifts timber from the area of felling to the main roads for further transportation. A new path-tracking algorithm is introduced, and demonstrated as being superior to standard techniques, such as Follow the Carrot and Pure Pursuit. To facilitate the research and development, a comprehensive software architecture for sensor and actuator interfacing is developed. Obstacle avoidance is accomplished by a new kind of radar, developed for and by the automotive industry. Localization is accomplished by combining data from a Real-Time Kinematic Differential GPS/GLONASS and odometry. Tests conducted on a simulator and a small-scale robot show promising results. Tests on the real forest machine are ongoing.