Ebook: Advanced Autonomous Vehicle Design for Severe Environments

Autonomy as a feature of ground vehicles is receiving broader and deeper research interest and engineering implementation in both unmanned vehicles and manned vehicle system applications. New directions in autonomous systems research and design have been gathering momentum during the past 10–15 years, including trajectory path planning and optimization, sensor data information fusion, vehicle kinematics and dynamics control, and many others. Even so, vehicle dynamic behavior and operational properties such as terrain and tactical mobility and energy efficiency, stability and safety, vehicle mission and task fulfillment in severe operational conditions (which include non-prepared terrain, climate and weather impacts, as well as today's aggressive cyber environments) have improved only incrementally at best. A detailed analysis showed that classical vehicle dynamics, which is the basis for manned ground vehicle design, has exhausted its potential to a large degree for providing novel design concepts. At the same time, unmanned ground vehicle (UGV) dynamics is still in its infancy and is currently developing using general analytical dynamics principles with very little involvement of actual vehicle dynamics theory.

In this regard, the main purpose of the NATO Advanced Study Institute (ASI) was to

(i) Analyze and present leading-edge research and engineering accomplishments and future trends in manned and unmanned ground vehicle dynamics as an analytical foundation for autonomous vehicle system design for severe environments and

(ii) Inspire young generation of researchers/engineers to work in these novel directions.

In keeping with these main goals, the ASI uniquely integrated renowned researchers from different areas of study whose contributions made it possible to establish new frontiers, methods and techniques in vehicle dynamics, severe terrain condition modelling, on-line control and learning, mechatronics and cyber-physics related issues. Accordingly, the book may be split into three parts: 1. novel research directions in vehicle dynamics that provide a basis for future autonomous system design; 2. advanced methods in on-line control and learning that can be utilized in autonomous system control design, and 3. main mechatronics and cyber-physics fundamentals to design for control, energy harvesting and cyber-security.

The book systematically presents material developed by the ASI's book contributors. Furthermore, the ASI inspired several contributors to continue their research work and thus the book includes new research results accomplished while the manuscript was being written after the ASI had been delivered.

The book opens with the historic aspects of ground vehicle dynamics and design, including major research done within the NATO. Building on that, it formulates and discusses future potential research directions in vehicle dynamics, such as inverse vehicle dynamics, agility in vehicle dynamics and agile mobility estimation, coupled and interactive system dynamics (i.e., fusion of dynamics of vehicle systems), advanced methods for wheel power distribution optimization to maximize energy/fuel efficiency and terrain mobility while minimizing damage of soil by the locomotion system.

Severe environment conditions and situations are presented in the book and advanced measurement, quantification and modelling methods are discussed. Also, an analysis of terrorist threats, agile vehicle trajectory deviation and critical reposition of a vehicle in interaction with environment is offered.

Mathematical and computational techniques and software products are described for the modelling of multibody vehicle dynamics in the presence of uncertainties, which are simulated with a polynomial chaos approach.

The on-line control, learning methods and techniques presented here bring together distributed and cooperative on-line control, reinforcement learning, and game theory to achieve a radical improvement in the agile performance of unmanned ground vehicles, including vehicle teams, at high speeds.

The book provides procedures for mechatronics design that are hierarchically separated into topological design and parametric design. The “mechatronics design for control performance” approach is explored and integrated into the overall design methodology and its optimization. Cyber-physics issues are detailed to emphasize all aspects of the vehicle mechatronics and ensure the security of the sensory communication system between the mechanical-electrical interconnected components.

Vehicle design and analysis for ride, handling and durability is addressed combining a description of how multibody systems analysis can be used to look at suspension systems, the use of finite element analysis to optimise lightweight vehicle structures and the influence of mass distribution on vehicle dynamics performance. The use of mathematical vehicle models and multi-objective programming to optimise suspension performance is also included and complements the use of standard software products.

This book could not be accomplished without the support and hard work of many people from various organisations. The editors and contributors would like to express their gratitude to the NATO Science for Peace and Security Programme which provided a grant to establish and conduct the ASI and to the reviewers who positively reviewed our ASI proposal and recommended it to NATO for approval. We thank all ASI attendees for their active participation, shared ideas and thoughts that have led to the successful completion of the ASI and, now, this book. Our sincere thanks to the ASI Organizing Committee including Heather Creel, Laura Harris, Rebecca Russell, Mostafa Salama, Sherrye Watson and many of our undergraduate and graduate students whose help with the ASI preparation was tremendous. We are grateful to the Faculty of Engineering and Computing at Coventry University who provided excellent facilities and hosted the ASI.

The editors look forward to continuing the research work and collaboration with the chapter contributors and express their hope for to share new finding in the near future.

Prof. Vladimir V. Vantsevich

University of Alabama at Birmingham, U.S.A.

Hoover, AL, U.S.A.

Prof. Michael V. Blundell

Coventry University, UK

Coventry, UK

This Chapter presents a brief analysis of research and engineering directions in ground vehicle dynamics and system design for more than 100 past years. It is shown that peculiarities of vehicle motion and dynamics of vehicles were established as an applied science – vehicle dynamics. Steady motion and transient maneuvers, multi-body dynamics, non-linear and stochastic dynamics, terramechanics, vehicle operational properties and their multi-criterion optimization, computer simulation, analysis and optimal synthesis, various types of control, inverse and direct vehicle dynamics approach, open architecture-type and multi-domain systems – all these makes the milestones of developments over the past century. New research directions were formulated in this Chapter, including agile interaction between ground vehicles and severe environments.

This chapter introduces classical and modern control theory by providing a short history of automatic control theory and describing basic elements and techniques of classical and modern control theory. Finally, the section concludes with a brief overview of existing autonomous platforms and navigation techniques in severe environments.

This section gives the reader an overview of the challenges related to dealing with the mobility and performance of vehicles in off-road conditions. It should be emphasized that the material of this section is important for both manned and unmanned vehicle applications. This is particularly relevant to the latter which operate in a wide range of severe terrains. The focus for such studies should be on the evaluation of the vehicle and operating environment as one interconnected system. Thus, a good understanding of soil mechanics is needed, as well as of the running gear of the vehicle.

The first part of the section is dedicated to the description of the indoor testing facility developed in the Advanced Vehicle Dynamics Laboratory (AVDL) at Virginia Tech. This test rig can be used with repeatable results to study the performance of pneumatic tires or rigid wheels on sand, soil, ice, or other type of surface. Illustrative examples from test studies performed on the rig on sandy loam and on ice are discussed. The second part of the section presents a study conducted at AVDL to compare the mobility, tractive performance, and energy efficiency of a robot in a wheeled and in a tracked configuration. The most significant results of this study are discussed.

It is not feasible to include in one section the state of the art in experimental terramechanics, tire and track modeling, and the challenges in assessing the mobility of robotic vehicles in off-road conditions. Dr. Sandu together with her students and collaborators performed significant research in all of these areas and others related in the last 15 years. The most significant publications that resulted from this work are included as references to this section. To ease the identification of studies dealing with a specific problem, the references were grouped as follows: The first group identifies publications resulted from Dr. Sandu's group in Experimental Terramechanics. As part of this group, [1–5] relate to Tire performance on ice, [6–8] relate to Tire performance on soil, [9–16] relate to Performance of robotic vehicles, and [17–19] relate to Test rigs design and modeling. The second group of publications is for Off-Road Vehicle Performance Modeling Studies. In this group, Tire modeling in off-road conditions occupies the largest section, [20–40], then Off-road vehicle performance modeling includes [41–45], Soil and terrain modeling is presented in [46–50], and Tracked vehicle performance in off-road conditions is included in [51–54]. The reader is encouraged to consult these references for a deep understanding of the experimental evaluation and mathematical modeling of the mobility, agility, and energy efficiency of manned and unmanned vehicles operating in unprepared conditions.

Security problems have emerged as soon as computers and networks entered our life and started to hold sensitive information. However network security of teams of vehicles still remains an open research problem despite its vital importance. This is because networked systems are difficult to observe and control even by expert users. Game theory provides an excellent framework to study network security problems because it captures the interlock between defensive and offensive algorithms. Network security can be viewed as a strategic game played between malicious attackers that manipulate data or compromise functionality and defenders whose aim is to protect information and maintain proper service operation. The security game framework is applicable to security problems in a variety of areas ranging from intrusion detection to social, wireless, and vehicular networks. This chapter introduces two different types of attacks (threats). Namely we consider measurement (vehicle trajectory deviation) and jamming attacks compromising the security of the networked teams. We propose two new techniques based on reinforcement learning to guarantee desired behavior in the presence of those attacks.

This section analyzes the transient period of the tire reaction force development process, which is characterized by the relaxation length, for the purpose of agile tire dynamics control as a pre-emptive, fast and exact response of a tire to dynamic changes of its interaction with terrain.

Researchers proposed and have used various approaches and functions for stochastic modeling of terrain. Usually, terrain topography profiles are modeled using the Monte Carlo method or other random processes. Such research methods are important for modeling dynamic normal reaction of tires and thus their influence on vehicle mobility, traction, ride and other operational properties. However, there is lack of methods on stochastic modeling of peak friction coefficient (also known as gripping coefficient) and rolling resistance coefficient. A new method for stochastic modeling of these two coefficients is considered in this section of the book. The method layouts a basis for agile/real-time terrain mobility, dynamics and mission fulfillment estimation of manned and unmanned ground vehicles.

A new method to estimate terrain mobility of a vehicle with a given number of wheels is presented. The method is founded on an estimation of mobility of each drive wheel. The method is based on stochastic terrain characteristics and agile tire slippage characteristics, which create fundamentals for agile tire slippage control and thus vehicle mobility enhancement.

The problem of overcoming a step obstacle by controlling the power distribution between the driving wheels of all-wheel-drive vehicles is analyzed. A mathematical model for traversing step obstacles by a single wheel that allows calculating the magnitudes of forces , including the torque and the longitudinal direction force that are needed to overcome the obstacle is developed. The results are applied to a 4x4 vehicle with individual distribution of power between the wheels.

A new method to estimate soil damage caused by vehicles' locomotion systems is proposed and includes the influence of the number of vehicle wheels, location of the wheels along the wheelbase, the distribution of normal load between the axles, driveline and tire characteristics.

A new method to improve vehicle mobility and energy efficiency of multi-wheel vehicles is proposed. It is shown the power distribution between the drive wheels impacts driveline power losses and power losses in the tire-soil contacts. Conditions for the maximum driveline efficiency analysed and the maximization of the tire slippage efficiency are analysed for a 6x6 terrain vehicle with different driveline configurations.

In this section, principles of wheel power split that are needed to design controls of all-wheel drive vehicles with hydrostatic drivelines are presented. Ecological aspects of the power distribution to the driving wheels are also included in the consideration and analyzed.

This section presents new inverse dynamics approaches that were established in the course of the author's vehicle dynamics research work. Case studies include an analysis of the influence of a driveline system on the power distribution between the drive wheels and optimisation of the gear ratio of the interaxle differential.

In this study we look into the treatment of parametric uncertainties and the response of a vehicle system with several degrees of freedom, using the polynomial chaos approach. The research aims at investigating the accuracy of the method compared with traditional Monte Carlo simulations, and the change in computational efficiency as the number of uncertain parameters and stochastic external excitations increase. A 7 degree-of-freedom full-car dynamic model has been developed. Key parameters of the system have been assumed to be stochastic, with large uncertainties. In addition, the vehicle runs over rough (and undeformable) terrain, with uncertain terrain height. The polynomial chaos expansion and the Galerkin approach are used to quantify the uncertainties and to determine the time evolution of the stochastic system under sole or combined sources of uncertainties.

We are not aware of any study trying to use the polynomial chaos framework to find a closed form solution for the LQR problem in this framework, i.e., a solution that would depend on the number of terms in the polynomial chaos expansions and that would numerically converge to the solution of the problem as S→∞. The original intent of the work presented here was to try deriving such a solution, but this proved to be extremely difficult, if not impossible. However, an efficient numerical method to solve this problem could be derived instead. Polynomial chaos based methods have the advantage of computationally much more efficient than Monte Carlo simulations. The method presented in this article treats the LQR problem as an optimality problem using Lagrange multipliers in an extended form associated with the polynomial chaos framework, and uses an iterative algorithm that converges to the optimal answer. Therefore, it goes at the root of the solution of the LQR problem, which is derived using Lagrange multipliers in the deterministic case, which leads to the well-known algebraic Riccati equations. Therefore, the method presented in this article might have the potential of being a first step towards the development of computationally efficient numerical methods for H_∞ design with parametric uncertainties.

In this study we employed a real-time estimation method for ground vehicle parameters based on a generalized polynomial chaos (gPC) approach applied to an extended Kalman filter (EKF) technique. The vehicle models considered were a load transfer model (LTM) and a modified load transfer model (MLTM). The data used for performing the parameter estimation was collected on a Land Rover Defender 110. Two sets of data were collected: one on a rural road and one on an urban road. The mass of the vehicle, as well as the lateral and longitudinal location of the center of gravity (CG), was estimated. The reason for developing the MLTM was that it has a significantly reduced need for prior knowledge about the CG location. The LTM requires that sensors be placed at the CG of the vehicle. Obviously, if that is already known, then this whole scheme is unnecessary. The MLTM is designed to remove that requirement through the ability to estimate the CG accelerations through the accelerometers placed at the four corners of the vehicle. The models and methods presented are validated against real data and parameters values with high accuracy. The computational cost is reasonable, and the estimator runs faster than real-time.

This Section looks at the use of industry standard multibody systems software for the modelling and simulation of vehicle dynamics. The MSC.ADAMS software is discussed as an example to demonstrate the principle modelling features used with a multibody systems approach. The modelling of rigid and flexible bodies is covered and attention is also given to the modelling of tracked vehicles, typical of those that are deployed to operate in severe environments. The modelling of tires is introduced but a more comprehensive treatment of this subject is provided in Chapter 9.

Section 8.1. Mechatronics is a synergistic application of mechanics, electronics, control engineering, and computer science in the development of electromechanical products and systems, through integrated design. A mechatronic system consists of sensors and transducers, actuators, and controllers. In developing a mechatronic system, modeling, analysis, integrated design, deployment, testing and refinement tasks are carried out. This section addresses Characteristics of a Mechatronic System, Motivation for Mechatronic Design, and Modeling Needs in Mechatronics.

Section 8.2. Mechatronics develops a unifying framework for modeling multi-domain (mechanical, electrical, thermal, fluid, etc.) systems which is capable of incorporating multi-functional devices into the framework. For this purpose, the concept of Through Variables and Across Variables are introduced in this section. An Across Variable is measured across the element. An Across Variable can be considered as Velocity, Voltage, Temperature, and Pressure in mechanical, electrical, fluid, and thermal systems, respectively. A Through Variable remains unchanged through the element. A Through Variable can be defined as Force, Current, Heat Transfer Rate, and Fluid Flow Rate for mechanical, electrical, fluid, and thermal systems, respectively.

Section 8.3. The objective of control is to make a (dynamic) system to behave in a desired manner, according to some performance specifications. All control systems involve a controller and a process. Control systems can be pneumatic, hydraulic, mechanical or electrical or a combination of them. Components of this block diagram are described below. In this section, Performance of control systems, and relation to Vehicle agility are discussed. Cyber physical systems for vehicles' control systems are also addressed in this section.

Section 8.4. A Self-powered Dynamic System is defined as a dynamic system powered by its own excessive kinetic energy, renewable energy or a combination of both. The technologies explored are associated with self-powered devices (e.g. sensors), regenerative actuators, and energy harvesting. The particular area of work is the concept of fully or partially self-powered dynamic systems requiring zero or reduced external energy inputs. The concept, theory and application of Self-powered Dynamic Systems is presented in this section. Optimal Uncertainty Quantification is introduced in this section as a powerful analysis tool.

Section 8.5. Mechatronic systems are complex and require multiple technologies in multiple domains. Sequentially designed components of existing “mixed” systems are not optimally matched. There is potential for improvement through concurrent and optimal design integrating several domains. This is the motivation for mechatronics design. Mechatronic Design Quotient (MDQ) approach is explained in this section as an efficient tool for design of mechatronic systems.

This Section introduces the subject of tire modelling, why tire models are important and how they are used in the vehicle dynamics subject area. Tire axis systems and geometry are also covered as are the methods used to test tires and gather the data needed to generate the parameters in a tire model.

This Section introduces the forces and moments that are generated in the tire contact patch with an emphasis on their representation in a vehicle handling model. Fundamental definitions such as slip ratio and relaxation length are provided. Example plots are shown to illustrate typical tire force and moment characteristics.

This Section introduces the area of tire modelling with a focus on the modelling of tire behavior to support the simulation of vehicles both on-road and off-road. For on-road work the emphasis is on empirical models, for off-road work the emphasis is on the physical modelling of the tire. The most recognized tire models in these areas, the Magic Formula and FTire are both also briefly introduced.

This Section covers the role of the suspension system for vehicle ride and handling and the design process involved with suspension systems to ensure good performance. The durability of the suspension system due to operating in aggressive environments and resilience to abuse loads is also discussed.

This Section covers the contribution of the vehicle body and the critical importance of producing light-weight structures with the correct mass distribution to ensure agile performance. The Dynamic Index design parameter and its influence on handling performance is described together with a practical example using an experimental vehicle. The use of finite element methods to optimize vehicle structures is also described.