Hybrid electric vehicles (HEVs) rely on energy management strategies (EMSs) to achieve optimal fuel economy. However, both model- and learning-based EMSs have their respective limitations which negatively affect their performances in online applications. This paper presents a computationally efficient adaptive dynamic programming (ADP) approach that can not only rapidly calculate optimal control actions but also iteratively update the approximated value function (AVF) according to the actual fuel and electricity consumption with limited computation resources. Exploiting the AVF, the engine on/off switch and torque split problems are solved by one-step lookahead approximation and Pontryagin's minimum principle (PMP), respectively. To raise the training speed and reduce the memory space, the tabular value function (VF) is approximated by carefully selected piecewise polynomials via the parametric approximation. The advantages of the proposed EMS are threefold and verified by processor-in-the-loop (PIL) Monte Carlo simulations. First, the fuel efficiency of the proposed EMS is higher than that of an adaptive PMP and close to the theoretical optimum. Second, the new method can adapt to the changed driving conditions after a small number of learning iterations and thus has higher fuel efficiency than a non-adaptive dynamic programming (DP) controller. Third, the computation efficiencies of the proposed AVF and a tabular VF are compared. The concise data structure of the AVF enables faster convergence and saves at least 70% of onboard memory space without obviously increasing the average CPU utilization.

Hybrid electric vehicles (HEVs) are irreplaceable in attaining sustainable development in contemporary society. Owing to the extra degree of freedom in supplying traction power, HEVs resort to appropriate energy management strategies (EMSs) to present their superiority over conventional internal combustion engine vehicles and pure electric vehicles.

Existing EMSs suffer from heavy computation overheads and excessive mode switches. This thesis proposes several novel methods for developing online EMSs for parallel HEVs that achieve both compelling fuel economy and excellent computation efficiency and adaptivity in online applications with uncertain driving conditions.

First, the solutions of offline dynamic programming (DP) are exploited to develop online EMSs for close-to-optimal control performances. The optimal speed profile serves as the reference in online control and the optimal value function (VF) is utilized to design control methods. To avoid the “curse of dimensionality”, the tabular VF is approximated by piecewise polynomials to substantially decrease the computation and memory overheads in online usage.

Second, to reduce the search space for optimal control actions, two types of special internal combustion engine (ICE) configurations are adopted and analyzed. The first type forces the ICE to strictly operate at the optimal operation line (OOL), whereas the second one allows a narrow band around the OOL. The second one outperforms the first one because it contributes to more robust ICE operations with slightly higher computation complexity.

Third, a hierarchical architecture is proposed for online EMSs so that the transient powertrain mode and torque split scheme are optimized by different methods in sequence. To avoid the exponential complexity of finding the optimal trajectory of the powertrain mode, the optimal VF is leveraged for an optimal decision within one sampling period with the aid of simplified assumptions. Model approximations on the ICE and the electric motor are conducted so as to convert the complex torque split problem into a constrained quadratic programming problem. These methods dramatically facilitate the computation efficiency of online EMSs.

Fourth, learning-based adaptive control is introduced to mitigate the adverse effect caused by the deviations between the model and the reality. For this target, efficient learning algorithms are designed to iteratively update the coefficient matrix of the approximated VF. Moreover, to avoid the pitfall of the “cold start” and prompt a fast convergence, the coefficient matrix is initialized by the optimal VF from offline DP.

Finally, an event-triggered control mechanism is applied to the torque split control and presents its remarkable advantage in eliminating the excessive computation overheads. At each time step, an efficient trigger algorithm decides if the reference ICE torque is still valid or outdated. If it is valid, the EMS directly uses the reference value as the optimal output; otherwise, the optimization algorithm for torque split control is executed to solve a new value and update the reference.

The performances of designed EMSs are tested by processor-in-the-loop simulations so that both the numeric results and the computation efficiency can be obtained for quantitative analysis and comparison. The testing results indicate that the designed EMSs can rapidly adapt to real driving conditions and generate more than 90% fuel economy of the DP optimum, and more importantly, all these EMSs can be implemented on a portable microprocessor with limited onboard computation resources.

Elhybridfordon (HEV) är oersättliga för att uppnå en hållbar utveckling i dagens samhälle. Medelst den extra frihetsgraden för att tillhandahålla dragkraft, använder HEV:erna sig av s.k. energihanteringstrategier (EMS) för att kombinera fördelarna med konventionella förbränningsmotorfordon och rena elfordon.

Befintliga EMS lider av tunga beräkningskostnader och överdrivna lägesomkopplare. Denna avhandling flera nya metoder för att utveckla online-EMS för parallella HEV som uppnår både övertygande bränsleekonomi och utmärkt beräkningseffektivitet, samt god anpassningsförmåga i online tillämpningar med osäkra körförhållanden.

För det första utnyttjas lösningarna från offline dynamisk programmering (DP) för att utveckla online EMS med nära optimal reglerprestanda. Den optimala hastighetsprofilen tjänar som referens vid online-reglering och den optimala värdefunktionen (VF) används för att utforma reglermetoder. För att undvika de problem som uppstår vid storskaliga beräkningar approximeras VF som en uppsättning polynom, vilket sänker bl.a. minneskostnader vid onlineanvändning.

För det andra analyseras två typer av speciella förbränningsmotorkonfigurationer (ICE) vilket minskar sökutrymmet för optimala regleråtgärder. Den första typen tvingar ICE att arbeta strikt vid den optimala driftlinjen (OOL), medan den andra typen tillåter ett smalt band runt den. Den andra typen är bättre än den första eftersom den bidrar till mer robust ICE-drift, men med något högre beräkningskomplexitet.

För det tredje föreslås en hierarkisk arkitektur för online-EMS så att det transienta drivlinjeläget och vridmomentfördelningen optimeras med olika metoder i sekvens. För att undvika den exponentiella komplexiteten i att hitta den optimala banan för drivlinjeläget utnyttjas en VF grundad på förenklade antaganden, vilket leder till ett optimalt beslut inom en samplingsperiod. Modellförenklingar av både ICE och elmotorn genomförs för att omvandla det komplexa problemet med vridmomentfördelning till ett begränsat kvadratiskt programmeringsproblem. Dessa metoder underlättar dramatiskt beräkningseffektiviteten för elektroniska EMS.

För det fjärde introduceras inlärningsbaserad adaptiv styrning för att mildra de negativa effekter som orsakas av avvikelserna mellan modell och verklighet. För detta mål utformas effektiva inlärningsalgoritmer för att iterativt uppdatera den approximerade VF. För att undvika en s.k. "kallstart" och för att få en snabb konvergens initialiseras VF med lösningen från offline-DP.

Slutligen tillämpas en händelseutlöst reglering av vridmomentfördelning, vilket påvisar anmärkningsvärda fördelar genom att eliminera överdrivna beräkningskostnader. Vid varje tidssteg avgör en effektiv utlösningsalgoritm om referensmomentet för ICE fortfarande är giltigt eller föråldrat. Om det är giltigt används det som referensvärde; i annat fall beräknar optimeringsalgoritmen för vridmomentfördelning ett nytt värde vilket uppdaterar referensen.

De utformade EMS:ernas prestanda testas med hjälp av simuleringar med beräkningar på dedikerad hårdvara (s.k. "Processor-in-the-loop"), så att både de numeriska resultaten och beräkningseffektiviteten kan erhållas för kvantitativ analys och jämförelse. Testresultaten visar att de utformade EMS:erna snabbt kan anpassa sig till verkliga körförhållanden och kan bidra till mer än $90\%$ bränslesparande jämfört med DP, och ännu viktigare är att alla dessa EMS:er kan implementeras på en bärbar mikroprocessor med begränsade beräkningsresurser.

Energy management strategies (EMSs) are crucial to the fuel economy of hybrid electric vehicles (HEVs). However, due to the lack of efficient solving approaches, most of existing EMSs mainly focus on the optimal torque split between the internal combustion engine (ICE) and the electric motor but neglect improper ICE on/off switches, and thus usually suffer degraded fuel economy and even unacceptable drivability in practice. To tackle this issue, this paper presents a novel EMS that uses an efficient actor-critic (AC) method to regulate ICE switches with limited computation resources. While common AC methods use complex neural networks (NNs) with arbitrary initialization, the proposed AC uses piecewise cubic polynomials whose parameters are initialized based on optimized solutions of dynamic programming (DP). By this means, the AC can quickly converge with high computation efficiency. The testing results from processor-in-the-loop (PIL) simulations showcase that, compared with a rule-based EMS with tabular value functions, the proposed EMS can greatly improve the equivalent fuel economy by eliminating improper ICE switches after only several iterations of adaptive learning and dramatically save onboard memory space owing to the concise AC structure.

The fuel economy of a hybrid electric vehicle(HEV) is determined by its energy management strategy (EMS), while the conventional EMS usually suffers from enormous computation loads when solving a nonlinear optimization problem. To resolve this issue, this paper presents a computationally efficient EMS with close-to-optimal performance using very limited computation resources. Relying on the optimal solutions by offline dynamic programming (DP), a constrained model predictive control (MPC) can quickly determine the engine on/off status and then the torque split problem is solved by a value-based Pontryagin’s minimum principle (PMP). Two measures are taken to further reduce the online computation cost: by surface fitting, the tabular value function is replaced by piecewise linear polynomials and thus the memory occupation is greatly reduced; and by model approximation, the nonlinear torque split problem becomes a quadratic programming one that can be more rapidly solved. The testing results from processor-in-the-loop (PIL) simulation indicate that the proposed EMS can generate a fuel efficiency close to the one by DP, but saves 70% onboard memory space and 30% CPU utilization compared with the benchmark EMS without taking the two measures.

To improve the fuel efficiency of a parallel hybrid electric racing car, this paper presents a computationally efficient energy management strategy (EMS) that regulates the internal combustion engine (ICE) to either operate with high efficiency or be switched off. Dynamic programming (DP) is employed to compute the optimal trajectories of vehicle speed and supercapacitor voltage. These trajectories are used as references for real-time online control. The online control contains two parts: a binary model predictive control (B-MPC) that determines the ICE on/off status and a Pontryagin’s minimum principle (PMP) that allocates the total torque demand to fuel and electric paths. The usual challenge to MPC is the large computational complexity and to PMP is the search of optimal costate The proposed EMS evades these two challenges. The binary property of MPC significantly reduces the computational complexity and the optimal costate of PMP is obtained from the value function computed by DP. The testing results by processor-in-the-loop simulation indicate the proposedEMS can realize a close to optimal performance and can be applied on a low-cost microprocessor.

This paper improves the fuel efficiency of a student-made parallel hybrid electric racing car whose internal combustion engine (ICE) either operates with peak efficiency or is turned off. The control to the ICE thus becomes a binary problem. Owing to the very limited computation resource onboard, the energy management strategy (EMS) for this car must have small time and space complexities. A computationally efficient controller that combines the advantages of dynamic programming (DP) and Pontryagins minimum principle (PMP) is developed to run on a low-cost microprocessor. DP is employed offline to calculate the optimal speed trajectory, which is used as the reference for the online PMP to determine the real-time ICE on/off status and the electric motor (EM) torques. The normal PMP derives the optimal costate trajectory through solving partial differential equations. The proposed quasi-PMP (Q-PMP) method finds the costate from the value function obtained by DP. The fuel efficiency and computational complexity of the proposed controller are compared against several state of art methods through both model-in-the-loop (MIL) and processor-in-the-loop (PIL) simulations. The new method reaches similar fuel efficiency as the explicit DP, but requires less than 1% onboard flash memory. The performance of the Q-PMP controller is compared between binary-controlled and continuously controlled engines. It achieves roughly 12% higher fuel efficiency for the binary engine with only approximately 1/3 CPU utilization.

Five existing definitions of autonomous driving test scenario were expounded, and the definitions of autonomous driving test scenario and related concepts were proposed by combining the logical relationships between test scenario, primitive scenario, and scenario elements. Three test scenario architectures of autonomous driving that are recognized in the industry were compared. From the perspective of scenario data sources, the current situation in the collection and research of traffic accident and naturalistic driving data at home and abroad was summarized. Based on the known data, expert data, test requirements, test objects, and technical characteristics of autonomous driving, the research results on the construction and automatic generation of unknown autonomous driving test scenarios were summarized. Research results show that the definition and architecture of autonomous driving test scenario are closely related to its construction and automatic generation. The autonomous driving scenario can be considered as an organic combination and comprehensive reflection of scenario elements, such as the driving environment, traffic participants, and driving behavior of autonomous vehicle. In addition to these elements, autonomous driving test scenarios should include a dynamic semantic description of initial state of the scenario, the situation of the scenario, and the impact and results at the end of the scenario. Although the existing test scenario architecture is relatively perfect, it is difficult to meet the requirements of different test objectives and test methods. Therefore, in the optimization of test scenario architecture, the design process of test scenario should be considered. However, the collection accuracy and effective characteristics of traffic accident data are not uniform, which makes it difficult to achieve the complete collection of naturalistic driving scenario data, and the collection specifications are not standardized. Therefore, the effectiveness of traffic accident and naturalistic driving data for the construction of autonomous driving test scenarios must be further demonstrated, and the autonomous driving test data are expected to become an important supplement. Improving scenario coverage and accelerating the testing process are important research goals in the construction of autonomous driving test scenarios. The in-depth application of artificial intelligence technology in the field of autonomous driving scenario generation is expected to meet complete or high coverage requirements of test scenarios. The classification of test scenarios for different levels of autonomous driving and the test scenario construction method for accelerated test of autonomous driving will be the next important research directions in test scenario construction for autonomous driving. 7 figs, 103 refs.

The fuel efficiency of the hybrid electric vehicle (HEV) highly relies on the engine efficiency. Thispaper proposes a particular powertrain configuration on a small hybrid electric racing car wherethe engine is only allowed to operate withpeak efficiency. An onlinebinarycontroller, accordingly, is developedbased on dynamic programming (DP)to control the engine on/off status and the torques from the electric motors (EMs) to ensure the HEV can successfully complete the drive mission with minimal fuel consumption.For comparison, the paper also develops anoptimal powertrain controller of the same HEV with the normal usage ofthe engine, i.e., the engine operates at any feasible point of the 2Dfuel efficiency map. The simulation results show that this binary controller can improveroughly13% on fuel efficiency compared with the general engine case.

To improve the fuel efficiency of a hybrid electric car with special powertrain features racing in Shell Eco-marathon, a computationally efficient online control system is developed by solving hierarchical optimal control problems. The top-level computes the optimal velocity trajectory based on the given competition track in advance. The lower-level then finds the best instantaneous engine state and torque allocation by the equivalent consumption minimization strategy (ECMS). The special design of the competition car reduces the ECMS into a binary optimization problem. The new controller can run in real-time on low-cost microprocessors and improves the car's fuel efficiency by 50% while maintaining the state of charge of the electrical energy buffer.

Hybrid electric vehicles (HEVs) rely on energy management strategies (EMSs) to achieve optimal fuel economy. However, both model- and learning-based EMSs have their respective limitations which negatively affect their performances in online applications. This paper presents a computationally efficient adaptive dynamic programming (ADP) approach that can not only rapidly calculate optimal control actions but also iteratively update the approximated value function (AVF) according to the actual fuel and electricity consumption with limited computation resources. Exploiting the AVF, the engine on/off switch and torque split problems are solved by one-step lookahead approximation and Pontryagin’s minimum principle (PMP), respectively. To raise the training speed and reduce the memory space, the tabular value function (VF) is approximated by carefully selected piecewise polynomials via parametric approximation. The advantages of the proposed EMS are threefold and verified by processor-in-the-loop (PIL) Monte Carlo simulations. First, the fuel efficiency of the proposed EMS is higher than that of an adaptive PMP and close to the theoretical optimum. Second, the new method can adapt to the changed driving conditions after a small number of learning iterations and then has higher fuel efficiency than a non-adaptive dynamic programming (DP)controller. Third, the computation efficiencies of the proposed AVF and a tabular VF are compared. The AVF data structure enables faster convergence and saves at least 70% of onboard memory space without obviously increasing the average CPU utilization.