In this paper, we introduce a novel method for safe navigation in agricultural robotics. As global environmental challenges intensify, robotics offers a powerful solution to reduce chemical usage while meeting the increasing demands for food production. However, significant challenges remain in ensuring the autonomy and resilience of robots operating in unstructured agricultural environments. Obstacles such as crops and tall grass, which are deformable, must be identified as safely traversable, compared to rigid obstacles. To address this, we propose a new traversability analysis method based on a 3D spectral map reconstructed using a LIDAR and a multispectral camera. This approach enables the robot to distinguish between safe and unsafe collisions with deformable obstacles. We perform a comprehensive evaluation of multispectral metrics for vegetation detection and incorporate these metrics into an augmented environmental map. Utilizing this map, we compute a physics-based traversability metric that accounts for the robot's weight and size, ensuring safe navigation over deformable obstacles.

Despite advances in integrating reasoning based on belief functions to generalise probabilistic representations, distance-to-prototype-based evidential deep neural networks are still emerging and require further consolidation. Existing studies in segmentation or classification tasks typically perform prior initialisation and do not address or mitigate the potential conflicts that may arise during fusion. This work investigates high-conflict scenarios within an evidential neural network for segmentation in autonomous driving, focusing on the distance-to-prototypes component, where prototypes, derived from feature maps, serve as sources of evidence and may yield contradictory information. Conflict is mitigated through parameter adjustments within the evidential reasoning, enhancing consistency before fusion. This enables more reliable data integration and a valid application of fusion rules and decision-making processes. The proposed rectification is validated on two prototype configurations of a deep evidential lidar-camera cross-fusion architecture, using two distance-based decision strategies and adapted metrics. The impact on the network's predictions is demonstrated through qualitative and quantitative results on road detection tasks with the KITTI dataset.

International audience

In self-navigation problems for autonomous vehicles, the variability of environmental conditions, complex scenes with vehicles and pedestrians, and the high-dimensional or real-time nature of tasks make segmentation challenging. Sensor fusion can representatively improve performances. Thus, this work highlights a late fusion concept used for semantic segmentation tasks in such perception systems. It is based on two approaches for merging information coming from two neural networks, one trained for camera data and one for LiDAR frames. The first approach involves fusing probabilities along with calculating partial conflicts and redistributing data. The second technique focuses on making individual decisions based on sources and fusing them later with weighted Shannon entropies. The two segmentation models are trained and evaluated on a particular KITTI semantic dataset. In the realm of multi-class segmentation tasks, the two fusion techniques are compared and evaluated with illustrative examples. Intersection over union metric and quality of decision are computed to assess the performance of each methodology.

Intelligent vehicles (IVs) are pursued in both research laboratories and industries to revolutionize transportation systems. Since the driving surroundings can be cluttered and the weather conditions may vary, environment perception in IVs represents a challenging task. Therefore, multi-modal sensors are engaged. In perception, outstanding performance is obtained by employing deep learning algorithms. However, deep learning often relies on probabilities while there is a better formalism to handle prediction uncertainty. To circumvent this, in this work, evidence theory is combined with a camera-lidar-based deep learning fusion architecture. The coupling is based on generating basic belief functions using distance to prototypes. It also uses a distance-based decision rule. Because IVs have constrained computational power, a reduced deep-learning architecture is leveraged in this formulation. In the task of road detection, the evidential approach outperforms the probabilistic one. Besides, ambiguous features can be prudently settled as ignorance rather than making a possibly wrong decision using probability. The coupling is also extended to the task of semantic segmentation. This shows how evidential formulation can be easily adapted to the multi-class case. Therefore, the evidential formulation is generic and produces a more accurate and versatile prediction while maintaining the trade-off between performances and computational costs in IVs. This work uses the KITTI dataset.

In this article, we propose a loose-coupling approach between the accurate Parallax Beam (PB) algorithm and an Inertial Measurement Unit (IMU) for a robust, off-road and system agnostic navigation application. With the aim of achieving better robustness against image-shortage and motion-estimation errors while keeping PB accuracy, we propose the use of the Multi-Sensor Fusion Extended Kalman Filter (MSF-EKF). The latter uses asynchronous data provided by both the IMU and PB with different sampling rates to perform a high rate estimation able to ensure correct navigation during image shortage. The evaluation performed on the KITTI vision benchmark suite shows promising results with a good accuracy and is able to run with a sampling rate ten times faster than the sampling rate of Parallax Beam.

Abstract Fifth-generation (5G) cellular mobile networks are expected to support mission-critical low latency applications in addition to mobile broadband services, where fourth-generation (4G) cellular networks are unable to support Ultra-Reliable Low Latency Communication (URLLC). However, it might be interesting to understand which latency requirements can be met with both 4G and 5G networks. In this paper, we discuss (1) the components contributing to the latency of cellular networks and (2) evaluate control-plane and user-plane latencies for current-generation narrowband cellular networks and point out the potential improvements to reduce the latency of these networks, (3) present, implement and evaluate latency reduction techniques for latency-critical applications. The two elements we detected, namely the short transmission time interval and the semi-persistent scheduling are very promising as they allow to shorten the delay to processing received information both into the control and data planes. We then analyze the potential of latency reduction techniques for URLLC applications. To this end, we develop these techniques into the long term evolution (LTE) module of ns-3 simulator and then evaluate the performance of the proposed techniques into two different application fields: industrial automation and intelligent transportation systems. Our detailed evaluation results from simulations indicate that LTE can satisfy the low-latency requirements for a large choice of use cases in each field.

Obstacle avoidance requires three main levels in autonomous vehicles, namely, perception, path planning and guidance control. In this paper, a global architecture is proposed by taking into account the link between the three levels. On the environment perception level, an evidential occupancy-grid-based approach is used for dynamic obstacle detection. The poses of objects are therefore considered for trajectory generation. The latter is based on a smooth trajectory sigmoid function. Finally, the control guidance employs this obstacle avoidance trajectory to generate the appropriate steering angle. The whole strategy is validated on our experimental test car. The experimental results show the effectiveness of the proposed approach.

Multiple Object Association is a considerable and challenging process in highly cluttered environments. Its aim is to thoroughly relate known objects to new detected ones which is hard in such conditions. The recurrent occlusions and pose variation of objects can raise ambiguity in decision-making. The objective of this paper is to establish a multi-feature fusion approach based on two sources describing, distinctly, the position and the motion direction of considered objects. The proposition is based on Dempster-Shafer theory to model the uncertainty and unreliability of sources and to ensure an evidential combination. Experimental results with real data from the KITTI database are presented to evaluate the proposed solution.

Vehicle-to-Everything (V2X) communication promises improvements in road safety and efficiency by enabling low-latency and reliable communication services for vehicles. Besides using Mobile Broadband (MBB), there is a need to develop Ultra Reliable Low Latency Communications (URLLC) applications with cellular networks especially when safety-related driving applications are concerned. Future cellular networks are expected to support novel latencysensitive use cases. Many applications of V2X communication, like collaborative autonomous driving requires very low latency and high reliability in order to support real-time communication between vehicles and other network elements. In this paper, we classify V2X use-cases and their requirements in order to identify cellular network technologies able to support them. The bottleneck problem of the medium access in 4G Long Term Evolution(LTE) networks is random access procedure. It is evaluated through simulations to further detail the future limitations and requirements. Limitations and improvement possibilities for next generation of cellular networks are finally detailed. Moreover, the results presented in this paper provide the limits of different parameter sets with regard to the requirements of V2X-based applications. In doing this, a starting point to migrate to Narrowband IoT (NB-IoT) or 5G - solutions is given.

Multiple Object Association (MOA) is an essential process in making vehicles smarter. It consists in associating newly detected objects to previously known ones at each instant in a scene. To treat such matter, the paper proceeds by the use of the Belief Theory introduced by Dempster-Shafer. A crucial concern in this application is the data association complexity and its relative computation time which must be distinguishably minimized. For that, this paper proposes a new approach in order to fit to real-time Intelligent Transportation Systems (ITS) applications. The solution consists in a first step on selecting the sources which deliver the most pertinent information and in the second step to combine the reduced data set. The effectiveness in the association quality as well as in the complexity reduction is shown with a pedestrian tracking example from the KITTI vision benchmark.

This paper focuses on the development of an IMU measurement simulator for navigation estimation algorithms validation. Its aim is to generate the sensor measurements thanks to an input trajectory described by the position and the orientation. The proposed models are derived from an inverse kinematic modeling of the sensors and an identification of their stochastic errors. These latter are composed of the biases instability, random walks and finally the sensors dynamics and bandwidth. The error model parameters of a low cost MEMS-IMU are determined using the Allan Variance method. In a second step, a Matlab simulator is built gathering the aforementioned models. Thanks to their completeness, this simulation tool is characterized by its wide range of application fields and dynamics that can be described. Its aim is to determine, from the time-dependent position and orientation data, the IMU measurements (3D accelerations and angular rates) without any object model. Finally, the simulator is validated using real experiments performed with an instrumented test car in normal driving as well as in obstacle avoidance situations.

This paper tackles the problem of tracking-based Traffic Sign Recognition (TSR) systems. It presents an integrated object detection, association and tracking approach based on a spatio-temporal data fusion. This algorithm tracks detected sign candidates in order to reduce false positives. Regions Of Interest (ROIs) potentially containing traffic signs are determined from the vehicle-mounted camera images. An original corner detector associated to pixel coding ensures the detection efficiency. The ROIs are combined using the Transferable Belief Model semantics. The associations maximizing the pairwise belief between the detected ROIs and ROIs tracked by multiple Kalman filters are processed. The track evolution helps to detect false positives. Thanks to this solution and to a feedback loop between the tracking algorithm and the ROI detector, a false positive reduction of 45% is assessed.

This paper presents an object tracking algorithm using belief functions applied to vision-based traffic sign recognition systems. This algorithm tracks detected sign candidates over time in order to reduce false positives due to data fusion formalization. In the first stage, regions of interest (ROIs) are detected and combined using the transferable belief model semantics. In the second stage, the local pignistic probability algorithm generates the associations maximizing the belief of each pairing between detected ROIs and ROIs tracked by multiple Kalman filters. Finally, the tracks are analyzed to detect false positives. Due to a feedback loop between the multi-ROI tracker and the ROI detector, the solution proposed reduces false positives by up to 45%, whereas computation time remains very low.

This paper describes a new strategy to assess a priori driving risk. The originality lies in the simultaneous consideration of the information related to the Vehicle, the Driver and the Environment (VDE). The heterogeneity and the imperfections of the information are taken into account thanks to the belief functions. A first fusion level determines the local risks related to the VDE entities, while their combination allows to determine the global a priori risk of the current driving situation. Both fusion levels are processed considering three combination rules to manage the eventual conflict. Comparative simulation results help to show the validity and the coherency of this multi-level risk assessment.

This paper describes a vehicle guidance strategy with a focus placed on the reference generation and the control levels. Further to a perception step, performed through the fusion of a Geographic Information System (GIS) and a vision system, the reference generation leads to the computation of a constrained smooth trajectory and a smooth speed profile integrating safety and comfort criteria. The obtained reference set is then used by a longitudinal and NLMPC-based lateral controller providing the steering angle and traction torque. The complete system performance are presented through simulation results based on real-time measurements.

This paper deals with a Speed Limit Determination Advanced Driver Assistance System (ADAS) performing the combination of a navigation system and a Speed Limit Sign Recognition System (SLSR). The present strategy is based on a multi-level data fusion using the Evidence theory. In a first step, the sensor reliabilities are estimated using indicators provided by the sources. Simultaneously, a multi-criterion fusion is processed on attributes extracted from the navigation system to detect its potential erroneous data and define the best navigation limit speed candidate. The second fusion level - the multi-sensor fusion - considers the navigation and the camera-based SLSR to be independent and specialized. In addition, the conflict is interpreted as an additional source of information for the final speed limit definition. The benefits of the proposed solution are shown through simulations and real experiments.

This paper presents a new constrained trajectory generation method dedicated to Advanced Driver Assistance Systems (ADAS). Based on the information provided by the digital map database of a navigation system, and considering different constraints related to the road profile, the vehicle and the driver, a convex optimization algorithm generates specific Spline-based trajectories. Characteristically, these trajectories are safe, they stay within the traffic lane borders, at the same time minimize an energy criterion along the path, and finally, they are curvature continuous. The present solution has been tested on several roads and the results show the efficiency of the energy constrained trajectory generation method.

In the context of aircraft, shimmy is an oscillatory phenomenon of the landing gear mainly due the tire dynamics and the landing gear structural dynamics. This phenomenon, which can result in severe structural damages of the landing gear, is here actively damped by a direct adaptive output feedback controller. The difficulties to model the ground/wheel interface require the use of an adaptive controller that can modify its behaviour in accordance with the plant dynamics. Thus, the proposed controller uses a fuzzy system to approximate the ideal feedback linearization law. Based on Lyapunov's theory, it is proven that the proposed adaptive control solution guarantees that all the error signals of the close-loop system are bounded. The main advantage of this control solution lies in the fact that the states of the system are not required for the synthesis of the controller. Simulation results show that the proposed control law creates a realistic control input which properly damps the oscillations. This work is supported by the European DRESS project (Distributed and Redundant Electromechanical nose gear steering System).