It is well acknowledged in the AI community that finding a good reward function for reinforcement learning is extremely challenging. Apprenticeship learning (AL) is a class of “learning from demonstration” techniques where the reward function of a Markov Decision Process (MDP) is unknown to the learning agent and the agent uses inverse reinforcement learning (IRL) methods to recover expert policy from a set of expert demonstrations. However, as the agent learns exclusively from observations, given a constraint on the probability of the agent running into unwanted situations, there is no verification, nor guarantee, for the learnt policy on the satisfaction of the restriction. In this dissertation, we study the problem of how to guide AL to learn a policy that is inherently safe while still meeting its learning objective. By combining formal methods with imitation learning, a Counterexample-Guided Apprenticeship Learning algorithm is proposed. We consider a setting where the unknown reward function is assumed to be a linear combination of a set of state features, and the safety property is specified in Probabilistic Computation Tree Logic (PCTL). By embedding probabilistic model checking inside AL, we propose a novel counterexample-guided approach that can ensure both safety and performance of the learnt policy. This algorithm guarantees that given some formal safety specification defined by probabilistic temporal logic, the learnt policy shall satisfy this specification. We demonstrate the effectiveness of our approach on several challenging AL scenarios where safety is essential.

Safety is a paramount concern when considering implementation of an automated highway where computers control the vehicles. Even with computer-fast reaction time there is inevitably some delay and if vehicles do not follow at safe distances, emergency braking maneuvers can cause dangerous collisions. This research investigates situations that might make automated vehicles have dangerous collisions and what standards the system design must hold to keep passengers safe.

The realm of autonomous robotics has seen impressive advancements in recent years, with robots taking on essential roles in various sectors, including disaster response, environmental monitoring, agriculture, and healthcare. As these highly intelligent machines continue to integrate into our daily lives, the pressing imperative is to elevate and refine their performance, enabling them to adeptly manage complex tasks with remarkable efficiency, adaptability, and keen decision-making abilities, all while prioritizing safety-aware navigation, mapping, and control systems. Ensuring the safety-awareness of these robotic systems is of paramount importance in their development and deployment. In this research, bio-inspired neural networks, nature-inspired intelligence, deep learning, heuristic algorithm and optimization techniques are developed for safety-aware autonomous robots navigation, mapping and control. A bio-inspired neural network (BNN) local navigator coupled with dynamic moving windows (DMW) is developed in this research to enhance obstacle avoidance and refines safe trajectories. A hybrid model is proposed to optimize trajectory of the global path of a mobile robot that maintains a safe distance from obstacles using a graph-based search algorithm associated with an improved seagull optimization algorithm (iSOA). A Bat-Pigeon algorithm (BPA) is proposed to undertake adjustable speed navigation of autonomous vehicles in light of object detection for safety-aware vehicle path planning, which can automatically adjust the speed in different road conditions. In order to perform effective collision avoidance in multi-robot task allocation, a spatial dislocation scheme is developed by introduction of an additional dimension for UAVs at different altitudes, whereas UAVs avoid collision at the same altitude using a proposed velocity profile paradigm. A multi-layer robot navigation system is developed to explore row-based environment. A directed coverage path planning (DCPP) fused with an informative planning protocol (IPP) method is proposed to efficiently and safely search the entire workspace. A human-autonomy teaming strategy is proposed to facilitate cooperation between autonomous robots and human expertise for safe navigation to desired areas. Simulation, comparison studies and on-going experimental results of optimization algorithms applied for autonomous robot systems demonstrate their effectiveness, efficiency and robustness of the proposed methodologies.