Combining machine learning and formal methods (FMs) provides a possible solution to overcome the safety issue of autonomous driving (AD) vehicles. However, there are gaps to be bridged before this combination becomes practically applicable and useful. In an attempt to facilitate researchers in both FMs and AD areas, this paper proposes a framework that combines two well-known tools, namely CommonRoad and UPPAAL. On the one hand, CommonRoad can be enhanced by the rigorous semantics of models in UPPAAL, which enables a systematic and comprehensive understanding of the AD system’s behaviour and thus strengthens the safety of the system. On the other hand, controllers synthesised by UPPAAL can be visualised by CommonRoad in real-world road networks, which facilitates AD vehicle designers greatly adopting formal models in system design. In this framework, we provide automatic model conversions between CommonRoad and UPPAAL. Therefore, users only need to program in Python and the framework takes care of the formal models, learning, and verification in the backend. We perform experiments to demonstrate the applicability of our framework in various AD scenarios, discuss the advantages of solving motion planning in our framework, and show the scalability limit and possible solutions.

Traditional supervisory control methods for the nonblocking control of discrete event systems often suffer from exponential computational complexity. Reinforcement learning-based approaches mitigate state explosion by sampling many random sequences instead of computing the synchronous product of multiple modular supervisors, but they struggle with limited reachable state spaces. A primary novelty of this study is to use the K-means clustering method for online inference with the learned state-action values. The clustering method divides all events at a state into the good group and the bad group. The events in the good group are allowed by the supervisor. The obtained supervisor policy can ensure both system constraints and larger control freedom compared to conventional RL-based supervisors. The proposed framework is validated by two case studies: an industrial transfer line (TL) system and an automated guided vehicle (AGV) system. In the TL case study, nonblocking reachable states increase from 56 to 72, while in the AGV case study, a substantial expansion from 481 to 3558 states is observed. Our new method achieves a balance between computational efficiency and nonblocking supervisory control.

Autonomous Intelligent Systems (AIS) are transforming various sectors by integrating advanced control theories, artificial intelligence, and cyber-physical systems. However, AIS control development faces significant challenges, including ensuring real-time responsiveness, designing adaptive controllers for dynamic environments, and coordinating multi-agent systems under uncertainties. These issues are exacerbated in resource-constrained settings, where balancing computational demands and real-time performance is critical.

To mitigate these challenges, this thesis leverages edge computing to enhance system performance, so that data-driven methods and optimal control technologies become feasible for complex AIS applications. Edge computing is a scheme that brings computation, communication, and storage resources closer to data sources, to achieve low-latency processing, real-time adaptability, and scalable solutions for AIS applications. It provides two key benefits: (1) offloading computationally intensive tasks to nearby edge servers, so as to ensure responsive and efficient operations despite constrained resources onboard, and (2) facilitating decentralized coordination among multiple agents by exploiting the edge server as a trustworthy node, so as to improve system scalability, reliability, and collaborative decision-making.

Building on the advantages of the offloading and coordination capabilities inherent in edge computing, this thesis investigates how these features can be harnessed to overcome the limitations of AIS in achieving optimal control and coordination. Primary contributions of this thesis include: (1) the development of state estimation and data-driven optimal control algorithms, which enables more precise estimation and control in nonlinear, time-variant systems; (2) the design of edge-based computational task offloading algorithms to achieve real-time adaptive control and learning by distributing computationally intensive tasks, which effectively balances latency and resource constraints; and (3) the introduction of decentralized optimization frameworks for multi-agent systems, which enhances scalability, robustness, and coordination under communication constraints by leveraging edge servers as trustworthy nodes for efficient collaboration and decision-making. All contributions have been validated through case studies in soft robotics and connected and autonomous vehicles, demonstrating their effectiveness and advancements over existing methods.

In summary, this thesis advances AIS capabilities by addressing real-time computational challenges and enabling optimal, data-driven control and decentralized coordination. The integration of edge computing improves the efficiency, scalability, and adaptability of AIS, offering promising opportunities for applications in autonomous mobility and other dynamic domains.

Autonoma intelligenta system (AIS) omvandlar olika sektorer genom att integrera avancerad reglerteknik, artificiell intelligens och cyberfysiska system. Utvecklingen av styrsystem för AIS står dock inför betydande utmaningar, såsom att säkerställa realtidsrespons, utforma adaptiva regulatorer för dynamiska miljöer och samordna multiagentsystem under osäkerheter. Dessa utmaningar är särskilt framträdande i resursbegränsade miljöer, där det är avgörande att balansera beräkningskrav och realtidsprestanda.

För att hantera dessa utmaningar utnyttjar denna avhandling edge computing för att förbättra systemprestanda och möjliggöra data-drivna metoder samt optimal styrning för komplexa AIS-applikationer. Edge computing innebär att beräknings-, kommunikations- och lagringsresurser flyttas närmare datakällorna, vilket möjliggör låglatensbearbetning, realtidsanpassning och skalbara lösningar. Detta tillvägagångssätt erbjuder två betydelsefulla fördelar: (1) avlastning av beräkningstunga uppgifter till närliggande edge-servrar för att säkerställa responsiva och effektiva operationer trots begränsade resurser ombord, och (2) decentraliserad samordning mellan flera agenter genom att utnyttja edge-servrar som tillförlitliga noder, vilket förbättrar systemets skalbarhet, robusthet och collaborative decision-making.

Genom att dra nytta av edge computing och dess möjligheter till beräkningsavlastning och decentraliserad samordning undersöker denna avhandling hur dessa funktioner kan användas för att övervinna AIS-begränsningar inom optimal reglering och koordination. De huvudsakliga bidragen i avhandlingen inkluderar: (1) utveckling av algoritmer för tillståndsskattning och data-driven optimal styrning, vilket möjliggör mer exakt skattning och styrning av icke-linjära, tidsvarierande system; (2) design av edge-baserade algoritmer för beräkningsavlastning, vilket möjliggör realtidsanpassad reglering och inlärning genom att distribuera beräkningstunga uppgifter och därmed balansera latens och resursbegränsningar; samt (3) introduktion av decentraliserade optimeringsramverk för multiagentsystem, vilket förbättrar skalbarhet, robusthet och koordination under kommunikationsbegränsningar genom att utnyttja edge-servrar som pålitliga noder för effektivt samarbete och beslutsfattande. Alla bidrag har validerats genom fallstudier inom mjukrobotik och uppkopplade autonoma fordon, vilket påvisar deras effektivitet och förbättringar jämfört med befintliga metoder.

Sammanfattningsvis stärker denna avhandling AIS genom att adressera realtidsberäkningsutmaningar och möjliggöra optimal, data-driven styrning och decentraliserad koordination. Genom att integrera edge computing förbättras AIS effektivitet, skalbarhet och anpassningsförmåga, vilket öppnar för lovande tillämpningar inom autonom mobilitet och andra dynamiska områden.

This study presents an innovative approach to optimal gait control for a soft quadruped robot enabled by four compressible tendon-driven soft actuators. Soft quadruped robots, compared to their rigid counterparts, are widely recognized for offering enhanced safety, lower weight, and simpler fabrication and control mechanisms. However, their highly deformable structure introduces nonlinear dynamics, making precise gait locomotion control complex. To solve this problem, we propose a novel model-based reinforcement learning (MBRL) method. The study employs a multi-stage approach, including state space restriction, data-driven surrogate model training, and MBRL development. Compared to benchmark methods, the proposed approach significantly improves the efficiency and performance of gait control policies. The developed policy is both robust and adaptable to the robot's deformable morphology. The study concludes by highlighting the practical applicability of these findings in real-world scenarios.

This study focuses on the locomotion capability improvement in a tendon-driven soft quadruped robot through an online adaptive learning approach. Leveraging the inverse kinematics model of the soft quadruped robot, we employ a central pattern generator to design a parametric gait pattern, and use Bayesian optimization (BO) to find the optimal parameters. Further, to address the challenges of modeling discrepancies, we implement a multi-fidelity BO approach, combining data from both simulation and physical experiments throughout training and optimization. This strategy enables the adaptive refinement of the gait pattern and ensures a smooth transition from simulation to real-world deployment for the controller. Compared to previous result using a fixed gait pattern, the multi-fidelity BO approach improves the robot’s average walking speed from 0.14 m/s to 0.214 m/s, an increase of 52.7%. Moreover, we integrate a computational task off-loading architecture by edge computing, which reduces the onboard computational and memory overhead, to improve real-time control performance and facilitate an effective online learning process. The proposed approach successfully achieves optimal walking gait design for physical deployment with high efficiency, effectively addressing challenges related to the reality gap in soft robotics.

The superiority of hybrid electric vehicles (HEVs) in energy efficiency highly relies on real-time executions of their energy management strategies (EMSs). Most current EMSs merely seek numeric optimality but neglect computation efficiencies in real-time applications. Hence, these EMSs usually suffer from tremendous computation overheads in practice and thus cannot be executed by onboard embedded processors. Consequently, this paper introduces an event-triggered control mechanism to replace the periodic torque split controller. When the HEV is in hybrid mode, an efficient trigger algorithm at each step determines whether the torque split controller needs to calculate a new solution or the previous solution is still valid. In this way, a large percentage of unnecessary computation overheads is avoided when the powertrain torque demand is unchanged. The advantages of this event-triggered control mechanism are demonstrated by processor-in-the-loop (PIL) simulations on different testing cycles. In contrast to the EMS with a fixed period, the event-triggered controller can significantly reduce both maximum and average CPU utilization in online testing without obviously compromising energy efficiency.

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.

Soft actuators have the advantages of compliance and adaptability when working with vulnerable objects, but the deformation shape of the soft actuators is difficult to measure or estimate. Soft sensors made of highly flexible and responsive materials are promising new approaches to the shape estimation of soft actuators, but suffer from highly nonlinear, hysteresis, and time-variant properties. A nonlinear and adaptive state observer is essential for the shape estimation from soft sensors. Current state estimation methods rely on complex nonlinear data-fitting models, and the robustness of the estimation methods is questionable. This study investigates the soft actuator dynamics and the soft sensor model as a stochastic process characterized by the Gaussian Process (GP) model. The unscented Kalman filter (UKF) is applied to the GP model for more reliable variance adjustment during the sequential state estimation process than conventional methods. In addition, a major limitation of the GP model is its computational complexity during online inference. To improve the real-time performance while guaranteeing accuracy, we introduce an edge server to decrease the onboard computational and memory overhead. The experiments showcase a significant improvement in estimation accuracy and real-time performance compared to baseline methods.

Modular supervisory control may lead to conflicts among the modular supervisors for large-scale discrete event systems. The existing methods for ensuring nonblocking control of modular supervisors either exploit favorable structures in the system model to guarantee the nonblocking property of modular supervisors or employ hierarchical model abstraction methods for reducing the computational complexity of designing a nonblocking coordinator. The nonblocking modular control problem is, in general, NP-hard. This study integrates supervisory control theory and a model-based deep reinforcement learning method to synthesize a nonblocking coordinator for the modular supervisors. The deep reinforcement learning method significantly reduces the computational complexity by avoiding the computation of synchronization of multiple modular supervisors and the plant models. The supervisory control function is approximated by the deep neural network instead of a large-sized finite automaton. Furthermore, the proposed model-based deep reinforcement learning method is more efficient than the standard deep Q network algorithm.

The decentralized approach is popular for the collaborative collision avoidance of connected vehicles in many scenarios. By modeling the task as a collaborative optimal control problem, Lagrangian methods are widely used to decouple the constraints and enable the decentralized solution. However, potential constraint-violating behavior will lead to oscillations during the Lagrangian update, resulting in more iterations and lower real-time efficiency. Existing methods generally neither address this shortcoming, nor explore the Lagrangian update mechanism. This study takes a control perspective, and solves this collaborative optimal control problem based on an extension of the Alternating Directions Method of Multipliers (ADMM) algorithm by performing the iteration update with a Proportional-Integral-(PI-) regulated controller. The link between the Lagrangian optimization and the PI controller improves the convergence performance during iterations. Simulation results in traffic intersection scenarios demonstrate the advantage of the proposed approach.