Current conventional lawn mowers have the following drawbacks; high initial costs, increasing engine noise levels, high operating costs due to high fuel consumption rates, the need to implement perimeter wires across the desired lawn mowing field, and high exhaustion of the operator in the long operating time. Hence, the need for a system that can achieve the same cutting effect of the existing lawn mowers with little or no operator’s fatigue, minimized noise pollution and running cost has risen. In present work, design of an autonomous solar lawn mower, a robotic vehicle that cut grass automatically with little human intervention, is discussed. The robotic vehicle is powered by Lithium-Ion batteries and depend on solar power delivered from a solar base station for charging the batteries. The mechanical design of the vehicle is flexible with the ability to control the height of the vehicle during grass mowing. Differential steering, with the implementation of multiple sensors is presented for obstacle avoidance. Also, a Raspberry Pi microcontroller for the image processing application through a camera is used for path planning. Furthermore, a solar based station structure is designed for charging the robotic vehicle batteries with minimized running cost, utilization of renewable energy source, no health hazards, not having any effect on the environment, and human’s effort and time are saved.

Advanced brake assist systems can avoid road accidents since the vehicles impact speed can be significantly reduced. To this end, different autonomous emergency braking systems are designed for recent vehicles on the market. This paper presents a pneumo-hydraulic Emergency Braking System (EBS) for autonomous racing vehicles. The purpose of the system is safely stopping the vehicle in case of any failures during autonomous driving. Failures can be detected both by the autonomous system itself and by human supervisors. The actuation system involves passive energy storage of compressed air to directly activate the hydraulic braking lines through pneumo-hydraulic pressure intensifiers. The coupling component between failures detection and actuation is a normally-open solenoid valve. The system is designed to respect deceleration and actuation time requirements, together with packaging constraints due to integration in an existing racing prototype. Specifically, the system requirements are specified by the racing competition rules: the overall reaction time of the retained EBS must be lower than 0.2 s, and the actuated mean deceleration must be greater than 8 m/s 2 on a dry track surface while keeping stable driving conditions. The validation and tuning of the system is performed in a simulated environment. Therefore, an extensive experimental validation of the system is required in the real applications.

This paper presents a local trajectory planning method based on the Rapidly-exploring Random Tree (RRT) algorithm using Dubins curves for autonomous racing vehicles. The purpose of the investigated method is the real-time computation of a trajectory that could be feasible in autonomous driving. The vehicle is considered as a three Degree-of-Freedom bicycle model and a Model Predictive Control (MPC) algorithm is implemented to control the lateral and longitudinal vehicle dynamics. The trajectory planning algorithm exploits a perception pipeline using a LiDAR sensor that is mounted onto the front wing of the racing vehicle. The MPC computes the acceleration/ deceleration command and the front wheel steering angle to follow the predicted trajectory. The trajectory and control algorithms are tested on real data acquisition performed on-board the vehicle. For validation purposes, the vehicle is driven autonomously during different maneuvers performed in the racing environment that is structured with traffic cones. The feasibility of the algorithm is evaluated in terms of error with respect to the planned trajectory, tracking velocity and maximum longitudinal acceleration. The effectiveness of the method is also evaluated with respect to command signals for the steering and acceleration actuators featured by the retained racing vehicle. The results demonstrate that the trajectory is well-tracked and the signals are compatible with the actuator constraints.

Visual simultaneous localization and mapping (vSLAM) and 3D reconstruction methods have gone through impressive progress. These methods are very promising for autonomous vehicle and consumer robot applications because they can map large-scale environments such as cities and indoor environments without the need for much human effort. However, when it comes to loop detection and optimization, there is still room for improvement. vSLAM systems tend to add the loops very conservatively to reduce the severe influence of the false loops. These conservative checks usually lead to correct loops rejected, thus decrease performance. In this paper, an algorithm that can sift and majorize loop detections is proposed. Our proposed algorithm can compare the usefulness and effectiveness of different loops with the dense map posterior (DMP) metric. The algorithm tests and decides the acceptance of each loop without a single user-defined threshold. Thus it is adaptive to different data conditions. The proposed method is general and agnostic to sensor type (as long as depth or LiDAR reading presents), loop detection, and optimization methods. Neither does it require a specific type of SLAM system. Thus it has great potential to be applied to various application scenarios. Experiments are conducted on public datasets. Results show that the proposed method outperforms state-of-the-art methods.

We describe a simulation environment that enables the development and testing of control policies for off-road mobility of autonomous agents. The environment is demonstrated in conjunction with the design and assessment of a reinforcement learning policy that uses sensor fusion and inter-agent communication to enable the movement of mixed convoys of human-driven and autonomous vehicles. Policies are learned on rigid terrain and are subsequently shown to transfer successfully to hard (silt-like) and soft (snow-like) deformable terrains. The enabling simulation environment is developed from the high fidelity, physics-based simulation engine Chrono. Five Chrono modules are employed herein: Chrono::Engine, Chrono::Vehicle, PyChrono, SynChrono and Chrono::Sensor. Vehicle’s are modeled using Chrono::Engine and Chrono::Vehicle and deployed on deformable terrain within the training/testing environment. Utilizing the Python interface to the C++ Chrono API called PyChrono and OpenAI Gym’s supporting infrastructure, training is conducted in a GymChrono learning environment. The GymChrono-generated policy is subsequently deployed for testing in SynChrono, a scalable, cluster-deployable multi-agent testing infrastructure built on MPI. SynChrono facilitates inter-agent communication and maintains time and space coherence between agents. A sensor modeling tool, Chrono::Sensor, supplies sensing data that is used to inform agents during the learning and inference processes. The software stack and the Chrono simulator are both open source. Relevant movies: [1].

Most controllers concerning lateral stability and rollover prevention for autonomous vehicles are designed separately and used simultaneously. However, roll motion influences lateral stability in cornering maneuvers, especially at high speed. Typical rollover prevention control stabilizes the vehicle with differential braking to create an understeering condition. Although this method can prevent rollover, it can also lead to deviation from a reference path specified for an autonomous vehicle. This contribution proposes and implements a coupled longitudinal and lateral controller for path tracking via model predictive control (MPC) to simultaneously enforce constraints on control input, state output, lateral stability, and rollover prevention. To demonstrate the approach in simulation, an 8 degrees of freedom (DOF) vehicle model is used as the MPC prediction model, and a high-fidelity 14-DOF model as the plant. The MPC-based lateral control generates a sequence of optimal steering angles, while a PID speed controller adjusts the driving or braking torque. The lateral stability envelope is determined by the phase plane of yaw rate and lateral velocity, while the roll angle threshold is derived from the load transfer ratio (LTR) and tire vertical force under the condition of quasi-steady-state rollover. To track the desired trajectory as fast as possible, a minimum-time velocity profile is determined using a forward-backward integration approach, subject to tire friction limit constraints. We demonstrate the approach in simulation, by having the vehicle track an arbitrary course of continuously varying curvature thus highlighting the accuracy of the controller and its ability to satisfy lateral and roll stability requirements. The MATLAB ® code for the 8-DOF and 14-DOF vehicle models, along with the implementation of the proposed controller are available as open source in the public domain.

Pedestrian-AV (automated vehicle) interaction attracts a significant amount of attention in recent years. However, there are controversial opinions about whether an external human-machine interface (eHMI) that communicates AV’s intent is needed. The authors propose that vehicle built-in cues (e.g., dynamic speed information) and the timing of displaying eHMI when communicating an AV’s yielding intent are both important. An online study was conducted to understand the impact on a 2-way street with a marked but not signalized crosswalk. As predicted, pedestrians were more likely to cross, with higher trust in AV and perceived safety, when the eHMI was displayed earlier. Information of the AV’s decreasing speed did not have a significant effect on crossing decisions, however, it increased trust and perceived safety. In addition, eHMI did have a positive add-on effect on pedestrian’s trust in AV, especially when it was displayed earlier, in the scenarios that dynamic speed of an AV was visible. Moreover, both the timing of eHMI and dynamic speed information played a role in enhancing a pedestrian’s judgment consistency in the short period of the interaction with AV, i.e., viewing the approaching AV from 20 meters away until it stopped near the pedestrian. The study indicated that both information about the vehicle’s physical behavior (i.e., speed) and the eHMI are beneficial in pedestrian-AV communication. More in-depth virtual reality or field studies are needed to further validate the findings.

Path tracking error control is an important functionality in the development of autonomous vehicles when a collision-free path has been planned. Large path tracking errors could lead to collision or even out of the control of the vehicle. Vehicle dynamic models are used to minimize the vehicle path tracking error so that control strategies can be designed under different scenarios. However, the vehicle dynamic model may not truly represent the actual vehicle dynamics. Furthermore, the nominal parameter employed in the vehicle dynamic model cannot represent actual operating conditions of the vehicle under environmental uncertainty. This paper presents a learning-based bias modeling method to improve the fidelity of any baseline vehicle dynamics model so that effective path tracking controller design can be achieved through a low fidelity but high-efficiency vehicle dynamic model with the aid of a few experiments or high fidelity simulations. The state-of-the-art of machine learning models, such as Gaussian process (GP) regression, recurrent neural network (RNN), and long short-term memory (LSTM) network, are employed for bias learning and comparison. A high-fidelity vehicle simulator, CARLA, is employed to collect virtual test data and demonstrate the effectiveness of the proposed bias-learning based control strategies under environmental uncertainty.

Shared autonomous vehicles (SAVs) encompassing autonomous driving technology and car-sharing service are expected to become an essential part of transportation system in the near future. Although many studies related to SAV system design and optimization have been conducted, most of them are focused on shared autonomous battery electric vehicle (SABEV) systems, which employ battery electric vehicles (BEVs) as SAVs. As fuel cell electric vehicles (FCEVs) emerge as alternative fuel vehicles along with BEVs, the need for research on shared autonomous fuel cell electric vehicle (SAFCEV) systems employing FCEVs as SAVs is increasing. Therefore, this study newly presents a design framework of SAFCEV system by developing an SAFCEV design model based on a proton-exchange membrane fuel cell (PEMFC) model. The test bed for SAV system design is Seoul, and optimization is conducted for SABEV and SAFCEV systems to minimize the total cost while satisfying the customer wait time constraint, and the optimization results of both systems are compared. From the results, it is verified that the SAFCEV system is feasible and the total cost of the SAFCEV system is even lower compared to the SABEV system. In addition, several observations on various operating environments of SABEV and SAFCEV systems are obtained from parametric studies.

Reliability-based mission planning aims to identify an optimal path for off-road autonomous ground vehicles (AGVs) under uncertain terrain environment, while satisfying specific mission mobility reliability (MMR) constraints. The evaluation of MMR during path planning poses computational challenges for practical applications. This paper presents an efficient reliability-based mission planning using an outcrossing approach that has the same computational complexity as deterministic mission planning. A Gaussian random field is employed to represent the spatially dependent uncertainty sources in the terrain environment. The latter are then used in conjunction with a vehicle mobility model to generate a stochastic mobility map. Based on the stochastic mobility map, outcrossing rate maps are generated using the outcrossing concept which is widely used in time-dependent reliability. Integration of the outcrossing rate map with a rapidly-exploring random tree (RRT*) algorithm, allows for efficient path planning of AGVs subject to MMR constraints. A reliable RRT* algorithm using the outcrossing approach (RRT*-OC) is developed to implement the proposed efficient reliability-based mission planning. Results of a case study verify the accuracy and efficiency of the proposed algorithm.

Operating system software of a prototype autonomous vehicle is presented in this paper. The prototype Autonomous Vehicle employs binary camera vision for navigation and is differentially steered. The software is a top-down design, modular in architecture to facilitate software maintenance and continual enhancement. At this present stage of development, it contains several parameter definition files and close to 50 major sub functions/sub function groups besides the main supervisory module. The subroutines range from system validation (data acquisition in real-time) and equipment calibration, image data filtering and processing to the driving and steering of the vehicle. The control codes are developed using the Borland Turbo-C compiler.

This paper presents a method for embedding human expert knowledge into a fuzzy logic controller. The method was developed while designing a fuzzy logic control system for an autonomous vehicle that utilized sparse sensor data for terrain classification. We will discuss the design for the fuzzy logic terrain classification system and use it as an example for explaining the embedding method. In the example, a human expert classified terrain features from sensor data provided by a two-dimensional computer simulation. From the information derived from the expert, we developed both a classification system for the terrain features and a fuzzy logic rule base for the controller. The simulation, along with an optimization algorithm, was then used to train the fuzzy logic controller to match the human responses.

This paper presents a control strategy that uses a hierarchical structure to arbitrate between recommendations from lower level modules. The lower level modules represent lower level tasks or behaviors. Each lower level module provides its own control recommendation based on its limited perception of the environment. We introduce the concept of a fuzzy quality measure that may be used by the hierarchical controller to determine how best to fuse the individual recommendations. The Quality Measure provides an approximate determination of each control recommendations potential value. The hierarchical partitioning reduces the cardinality of the rule base and decreases the number of system parameters, as compared to a monolithic structure. Optimization of the reduced parameter set is simpler and requires less time.

Autonomous vehicle (AV) is considered to be a significant part of the future of the automotive industry, with advantages such as higher transportation safety, less greenhouse gas emission, and wider transportation accessibility. However, the fruit of this technology cannot meet with much success in the market without an appropriate level of human trust, which might lead to refusing to adopt it, or using it in a wrong way. Therefore, except for improving safety of the technology itself, interactions between AV and human around it, such as pedestrians, are also paramount, especially effective communication. It is expected that certain communication could improve trust, and enable people to respond and act towards it accurately. This paper conducted a preliminary online study for a proof of concept that communication style (informative, advisory, and commanding) and communication strategy (single communication, adaptive communication) could affect pedestrian crossing decision process. The results demonstrate that advisory style improves trust in AVs when a Cross decision is expected by AVs, while both commanding and advisory styles improve trust in AVs when a Not Cross decision is expected by AVs. In addition, adaptive communication strategy improves trust significantly when AVs expect pedestrians to cross. This study helps understanding the communication between AV and pedestrians and is instrumental in developing more advanced AV communication mechanisms.

Connected autonomous intelligent agents (AIA) can improve intersection performance and resilience for the transportation infrastructure. An agent is an autonomous decision maker whose decision making is determined internally but may be altered by interactions with the environment or with other agents. Implementing agent-based modeling techniques to advance communication for more appropriate decision making can benefit autonomous vehicle technology. This research examines vehicle to vehicle (V2V), vehicle to infrastructure (V2I), and infrastructure to infrastructure (I2I) communication strategies that use gathered data to ensure these agents make appropriate decisions under operational circumstances. These vehicles and signals are modeled to adapt to the common traffic flow of the intersection to ultimately find an traffic flow that will minimizes average vehicle transit time to improve intersection efficiency. By considering each light and vehicle as an agent and providing for communication between agents, additional decision-making data can be transmitted. Improving agent based I2I communication and decision making will provide performance benefits to traffic flow capacities.

While there are many factors that affect traffic accidents, safe road design plays a key role. It is critical to understand traffic flow and driver response to various scenarios. Furthermore, with the inclusion of new traffic trends such as bicycle friendly roads and autonomous vehicles, it is ever more important to study the complex relationships between traffic entities. However, controlled experiments can be difficult to implement in the real world. Traffic simulation has been researched for decades to understand the flow of traffic in a variety of environments and how users interact within it. However, these platforms are limited in simulation and authoring capabilities, do not allow for multiple human agents, or are not accessible by roadway designers. This research explores the technical development of a multi-user multi-modal traffic simulation platform that expands on the capabilities of traditional traffic simulators. Traffic simulation of virtual cars and pedestrian agents were generated through VISSIM and networked into a simulation platform. Multiple human agents can interface with the system through a physical car and bicycle rig featuring both a virtual reality head mounted display and a multi-monitor display. Furthermore, the multi-user nature of the platform allows for a variety of complex research applications including multi-user behavior. By utilizing low-cost commodity hardware and software, the accessibility of this platform is greatly increased. A testbed environment was created to evaluate the technical limitations of the platform and assess where further work is needed. Through this evaluation, the authoring process and networking capabilities were optimized. Additionally, the core functionality of the platform, including the various simulator modes and multi-user functionality, was found to be successful. Through this research, combining commodity hardware with traffic simulation will allow a larger number of researchers to better understand traffic environments and how humans may interact within them.

This paper presents a method based on a Model Predictive Control (MPC) aiming to optimize the passenger comforts in assisted and autonomous vehicles. The controller works on the lateral and longitudinal dynamics of the car, providing front wheel steering angle and acceleration/deceleration command. The comfort is evaluated through two indexes extracted from the ISO 2631: an equivalent acceleration a eq and a Motion Sickness Dose Value (MSDV) index. The MPC weighting parameters are designed according to the values assumed by these indexes. Specifically, each weighting parameter is changed until the most satisfying comfort evaluation and the maximum vehicle performances, in terms of lateral deviation, tracking velocity and relative yaw angle, are reached. The controller is tested numerically on a simulated scenario resulting from real GPS data obtained in a highway. The method is compared with an alternative control strategy based on the combination of a PID and a Stanley control for the longitudinal and lateral dynamics, respectively. The results demonstrate the effectiveness of the approach, leading to a low percentage of passengers can experience motion sickness.