Autonomous driving is a highly topical research area, where significant positive impacts on safety and environment can be made, especially in the trucking industry. The vehicles in this industry often consist of a tractor unit combined with a trailer. This project focuses on navigating a model semi-trailer truck through an urban- like environment. A number of challenges arise from these settings, such as path planning and control through sharp turns and crossings, combined with obstacle avoidance. This needs to be done with high precision, considering that the whole articulated vehicle needs to stay within the bounds of the road. Since the vehicle will need to take critical decisions quickly, the performance and reliability of the control system is also important.
are unobservable. Privara highlighted three main challenges in data-driven modeling for MPC: 1) data violates typical persistent excitation requirements, 2) increased model complexity increases length of learning time and suitable experiments may be expensive, 3) measured temperature signals are often co- linear . As detailed throughout the adaptive control literature, without suf- ficient excitation, catastrophic parameter estimate drift may occur [35, §5.2 & §8.3]. Excitation must be sufficient to achieve the signal to noise ratio and infor- mation content required for parameter estimation. Active excitation techniques can reduce plant/model mismatch by using input signals to actively excite and perturb the system. System identification techniques design excitation signals using sum of sinusoids or pseudo-random binary inputs , , , . Similarly, extremum seeking uses a dithering signal to estimate the objective function gradient . In general these brute force excitation techniques can be costly and disruptive to the desired system operation. As a result, optimal ex- periment design seeks to minimize model uncertainty and excitation power or energy .
Currently, the literature on differential-game-based autonomous vehicle control can be categorized into two categorries with respect to the scenario of the study. The first one is autonomous racing [138, 139, 140]. Both vehicles in the game want to reach the destination ahead of the other. The second one is collision avoidance in the motion planning problem [141, 142, 140]. Both players need to reach its own destination, say traveling across unsignalized intersection, without colliding with the other. In addition to the above two categories, Huang uses mean field game  to study the behavior of mass behaviors instead of individual agents in traffic flow. Wang et al. apply cooperative differential game to car-following and lane change model . Lygeros et al. study the platooning behavior of two vehicles from the perspective of zero-sum game. However, few literature models general highway driving scenario as non-cooperative differential games.
capabilities. It can be combined with various system models, control theory and optimization algorithms to form various control algorithms. MPC can overcome tracking errors and random disturbances in complex environments, and eliminate the uncertainty caused by model mismatch and external disturbances in time; it is also suitable for systems in which mathematical models are inaccurate and have constraints. Consequently, MPC is widely used in the motion control of smart cars, drones and mobile robots  and . Shen et al. performed trajectory tracking control of autonomous underwater vehicle using Lyapunov-based model predictivecontrol . Falcone et al. applied MPC to an active front steering system in an autonomous vehicle . Based on the vehicle dynamics model, Lee et al.  and Kim et al.  applied the MPC method to laterally control a vehicle and achieved good control results. Researchers studied the trajectory control method of a tractor-trailer system based on MPC, and the tests proved that the control method is effective  and . Therefore, applying the model predictivecontrol method to the electric drive track vehicles can achieve good control effects, but at this stage there is less research in this direction.
This paper develops a linearized time variant model predictivecontrol (MPC) approach for controlling autonomous vehicle tracking on feasible trajectories generated from the vehicle nonlinear ordinary differential equations (ODEs). The paper is an application of the results from computational schemes for nonlinear model predictivecontrol published in (International Journal of Control, Automation and Systems 2011 9(5), 958- 965; Mechatronics 2013, Trajectory Generation for AutonomousVehicles, 615-626, Springer). The vehicle nonlinear dynamic equations are derived and solved in MPC optimizer. Solution for the closed loop control is obtained by solving online the vehicle dynamic ODEs. Simulations for the new schemes are presented and analysed.
Until now, the longitudinal dynamics control of the vehicle has largely remained un- der the full authority of the driver being restricted in systems such as the CC for comfort reasons and in autonomous vehicle control applications. In addition, braking systems for DYC that decelerate the vehicle are mostly viewed as depreciating on the driving expe- rience [46, 54]. Although it is well known that the driver should remain at the centre of the longitudinal dynamics control, later research has proven that active control can im- prove the stability in limit handling situations [24, 39]. Terminal understeer arises when an overspeeding vehicle enters a turn and its turning radius cannot be decreased to match the minimum turn-radius given by its velocity and understeer gradient. One of the earliest studies which explored this idea is  where they noted that stability and path tracking is improved with the combination of a corrective yaw moment and braking through appro- priate brake control of the four wheels. More recently, Rajamani and Piyabongkarn  concluded that a reduction of lateral acceleration by decreasing the velocity of the ve- hicle before entering a sharp turn provides a better cornering performance and rollover mitigation than a typical yaw rate controller. Reduction of lateral acceleration results in reduction of slip angle at the tyres and lower chances of exceeding the limit of adhesion.
A company that has recently developed a new Autonomous Terminal Tractor (ATT) approached Distribute and asked for insights in what kind of equipment the ATT would need. An ATT is an autonomous vehicle that is newly developed and can be used in container terminals to transport containers. The hardware of the new ATT is finished, but the software and the equipment are yet to be installed. To make the ATT as attractive as possible for potential buyers, it should be able to drive efficiently and smart. Therefore, it should be equipped with the right sensors, cameras and logic. While the ATT will eventually be used in container terminals, it is important to look at the properties of this environment. The most important goal of a container terminal is to load or unload vessels as quickly as possible to be able to guarantee a short VTT. The task of ATTs in this process would be delivering the right containers on the right time. Mainly, three jobs can be distinguished in a container terminal and these are shown in Table 1.
The evolving class of electric vehicles (EV) with near-wheel motors and brake-by-wire provides new possibilities of motion control, such as torque vectoring and hybrid braking. Our robotic vehicle ROMO , with four in-wheel motors (IWM), represents an example of such powertrains. In these powertrains, the regenerative capabilities of the traction motors can be used to support the brake-by-wire friction brake (FB) system during braking. In this situation the motor is used as a generator and recharges the batteries or capacities of the EV. Therefore, from an energy-consumption standpoint, it is preferable to maximize the usage of the motor torque during braking. Moreover, due to the fast and precise response of the traction motor, this electric actuator can also be used to improve the bandwidth of the wheel torque and wheel slip control .
The difference between these different control loops is the decision making cycle. There is no argument about sensing and acting as critical stages in the process, the question remains, how does a robot decide what action to take given particular sensor data and user inputs? For this there are two distinct schools of thought. One framework follows a probabilistic approach, where feature tracking is achieved by having the machine learn hidden Markov model chains (HMM) from past AUV runs. A large number of scenarios are needed in order to train the HMM to make good estimates of the current environment state. When in a real mission the vehicle will act upon the most likely scenario, as estimated by the HMM. A second framework looks at the problem from an optimization perspective, where for a given set of observations the system reaches a decision that maximizes the utility function. Given that there are a number of possible behaviours this is often a multiple objective optimization problem – different approaches are proposed to solve this problem.
Training an RNA is an essential process to reach the desired goal. Currently there are many platforms that allow this task from a personal computer, although this does not mean that with a single processor we can solve RNA with thousands of neurons. For the training of the network, the Matlab software was used, which provides many functions and tools that allow us to train and see the behavior of our RNA . In addition, an Excel sheet was used to compare the data obtained by the Matlab software. Subsequently, the necessary code was developed in C ++ to be loaded in autonomousvehicles.
The purposes of such vehicles are extremely various, ranging from scientific exploration, data collection and remote sensing, provision of commercial services, military reconnaissance and intelligence gathering. Recently, unmanned systems have become available and research is ongoing in a number of areas that will significantly advance the state of the art in unmanned vehicles technology. Moreover, designers have more freedom in the development of such vehicles, not having to account for the presence of a pilot and the associated life-support systems. This potentially results in cost and size savings, as well as increased operational capabilities, including fault diagnosis and fault tolerant supervision systems [1–5].
changing lanes) . Recently, V2V communication have pushed the ACC system into a more sophisticate system, called CACC. Each vehicle within the cooperative driving system is equipped with on-board sensors measuring position, velocity, acceleration. Such set of measurements requires Inertial Measurements Units (IMU), Global Positioning Systems (GPS) and radars, which are commonly available on road vehicles. Each vehicle is also equipped with wireless V2V communication hardware to share information with its neighbors and receive reference signals. Thanks to the information of neighbors vehicles CACC controller will be able to anticipate problems better, enabling it to be safer, smoother and more reliable in response. In CACC, wireless communication is used by the controller to regulate speed and distance between vehicles, ensuring that any changes in speed by the driver in front of you are immediately registered in the cooperative vehicle. However, most of the CACC controller presented in literature does not cope communication failure/impairments, network delay and security vulnerabilities. To over- come these issues, flexible control system, reconfigurable on the basis of the actual communication capabilities, have to be designed. In this sight, cooperative driving can be represented as a networked control system where the vehicles are controlled by handling their state information and networked information received from neighboring vehicles through the communication network [170, 167, 144] in which the time-delay and the security vulnerabilities are explicitly modeled in order to give a more realistic representation of the cooperative driving systems [151, 155, 152].
To set the scene for the extent of coverage required: initial operations extending at least over the first year were planned to be at inshore locations, launch and recovery were to be from a pier using a dockside crane, and the AUV operations were to be within 5km and line of sight of the launch point. Members of the core operating team are university employees, not students, who have attended training in operation of the AUV with the manufacturer, as well as gaining other experience with well known experienced AUV operating teams. The AUV will be followed during its missions by a tracker boat that will remain in radio contact with the AUV operations control station. The tracker boat will fly a “diver/equipment in the water” flag. The operations are to be in coastal Newfoundland where inshore traffic is relatively low.
According to the United Nations Environment Program (UNEP) report , district energy systems, such as the virtual power plant (VPP), can create a pathway to transit from intense use of fossil fuel and achieve a 30–50% reduction in primary energy consumption by harnessing renewable resources. This is also in line with regional mandates such as the European Commission’s 20-20-20 strategic objective to increase renewable uptake by 20% by the year 2020 and ensure a low-carbon economy by 2050 [3,4]. Although renewable uptake is increasingly explored , its sources are intermittent in nature, which can lead to voltage fluctuations and loss of loads. This can be compensated by energy storage systems  including battery banks in plug-in electric vehicles (PEVs). VPP provides a platform for smart coordination of distributed energy sources and loads in geographically-dispersed environments such as educational campuses, industrial parks and small communities. By integrating
actuators, see e.g. [1–4]. A CA method for an over-actuated vehicle uses a modular architecture for the control where a high-level motion controller computes the resulting forces and moments on the vehicle requested to cope with a specific manoeuvre. Once these forces and moments are known, they are sent to a low-level coordination controller, where CA is implemented, to find a suitable use of the motion actuators so that they produce the requested resulting forces and moments on the vehicle. CA methods usually do not consider the dynamics of the motion actuators, possibly only rate limitations. On the other hand, as we are trying to coordinate motion actuators with different behaviours, it is useful to introduce in the controller an explicit formulation of the actuators dynamics by using a Model PredictiveControl (MPC) approach. The resulting control structure, referred to as Model PredictiveControl Allocation (MPCA) in the following, has recently been used in other fields of research. Within aerospace [5–7] use an MPCA approach for the inner loop of a re-entry vehicle guidance and control system, while in  as part of a missile flight control system. In the automotive area, [9, 10] propose the MPCA strategy to control an engine thermal management system, while  use it for the hybrid braking of an electric vehicle with four in-wheel motors. The performance of a vehicle when facing the three proposed scenarios, split-µ braking, split-µ acceleration and brake blending, has never been studied using MPCA. Moreover, no documentation has been found about the implementation of MPCA to improve the dynamics of a heavy vehicle. The objective of this paper is to investigate if MPCA can cope with the three above mentioned scenarios and compare the performance achieved by MPCA with CA. The article is structured as follows. Section II starts with a brief introduction of CA and MPC methods, followed by a description of the designed MPCA together with the solver used during performed simulations and tests. Section III describes in detail the three selected scenarios while section IV shows the results from both simulations and tests. The conclusions, benefits and limitations of the proposed MPCA are stated in section V. Variables and signs are in accordance with ISO 8855:2011  and units are expressed according to SI, unless otherwise specified.
(2). Automobile plays an important role in modern society. However, with the economic benefits and living convenience provided by automobile, traffic accidents also become an important factor to cause personal and property losses. Many traffic accidents could be avoided if vehicles were given the ability to predict and deal with dangerous situations by themselves. Pedestrian detection based on video image data and corresponding countermeasures are obviously an important part of realizing this kind of vehicle assisted driving ability. At present, many companies and academic institutions made many relevant researches, such as Google, Tesla, MIT and baidu. With these researches getting deep, the demanding for vehicle assisted driving technology is becoming stronger and stronger, which has been a hot issue of common concern of academia and industry. As an important part of vehicle assisted driving technology, pedestrian detection algorithm has made some progress, but its ability to face complex scenes is still a problem.
Singapore’s Traffic Police shows that there are many motorists with dangerous driving habits. There was continued increase in the number of speeding and red-light running violations. In 2012, 2,917 people were arrested for drink-driving (Channel NewsAsia 2013), and there were 168 people killed and 9,106 injured in road accidents in Singapore (Singapore Police Force 2013). In self-driving vehicles, irresponsible driving behaviour and human errors in driving would be eliminated. With fewer road accidents, there would be less traffic jams, injuries and fatalities, lower medical costs associated with accidents, fewer insurance claims and hence lower premiums. Car rides could be less stressful for the drivers, who would instead spend their time on other activities in the cars.
A technical project paper entitled “The Robotic Arm”  by Yu Shan Zhen, Li Feng and Randall Watanabe, FALL (1997) is also studied. The objective of this design project is to use a XILINX FPGA chip, XC4010E, to build a Controller System to control the movements of the robotic arm. The whole system is composed of the Controller System and three drive circuits. One driver circuit for each motor on the robotic arm. The Control System will feed the drive circuits that actually drive the motors on the robotic arm. These drive circuits are needed because the Control System does not supply enough power to drive the motors directly. The controller System is implemented on the XC4010E XILINX chip. It has two inputs and six outputs. One of the inputs is a reset switch that resets the Control System to the initial state. The other input is an external clock used to synchronize the output signals. It will be a 1 kHz signal generated with a signal generator. The XILINX FPGA is capable of running at much higher speed but a slow clock is needed to obtain relatively large delays for the output signals. The six output signals form three pairs. Each pair of signals is for each motor on the robotic arm. Since there are three pairs, there are three motors on the robotic arm, one for up and down movement, one for left and right movement and another for grasping and ungrasping. The drive circuits are built with TTL Logic gates and NPN transistor amplifiers, where the Logic gates ensure the proper input into the transistors. To verify the movement of the robotic arm, the programmer need to do try and error type of programming to ensure the movement of the robot, this will take some time to verify the movement of the robot.
Electric vehicles are expected to become one of the key elements of future sustainable transportation systems. The first generation of electric cars are already commercially available but still, suffer from problems and constraints that have to be solved before a mass market might be created. Key aspects that will play an important role in modern electric vehicles are range extension, energy efficiency, safety, comfort as well as commu- nication. An overall solution approach to integrating all these aspects is the development of advanced driver assistance systems to make electric vehicles more intelligent. Driver assistance systems are based on the integration of suitable sensors and actuators as well as electronic devices and software-enabled control functionality to automatically sup- port the human driver. Driver assistance for electric vehicles will differ from the already used systems in fuel-powered cars such as electronic stability programs, adaptive cruise control etc. in a way that they must support energy efficiency while the system itself must also have a low power consumption. In this work, an eco-driving functionality as the first step towards those new driver assistance systems for electric vehicles will be investigated. Using information about the internal state of the car, navigation informa- tion as well as advanced information about the environment coming from sensors and network connections, an algorithm will be developed that will adapt the speed of the vehicle automatically to minimize energy consumption. From an algorithmic point of view, a stochastic model predictivecontrol approach will be applied and adapted to the special constraints of the problem. Finally, the solution will be tested in simulations as well as in first experiments with a commercial electric vehicle in the SnT Automation & Robotics Research Group (SnT ARG).
embraced AV technology, and this coverage discrepancy highlights an area of the Act that should be monitored going forward, especially once the insurance liability coverage minimums revert to their original amounts. In the future, the legislature may need to consider crafting new legislation in the context of fully autonomous vehicle insurance coverage, rather than basing it on existing limousine and taxi law. 134