Abstract-- Quadrotor motion planning is critical for autonomous flight in complex environments, such as rescue operations. Traditional methods often employ trajectory generation optimization and passive time allocation strategies, which can limit the exploitation of the quadrotor's dynamic capabilities and introduce delays and inaccuracies. To address these challenges, we propose a novel motion planning framework that integrates visibility path searching and reinforcement learning (RL) motion generation. Figure 1: Illustration of the proposed method. Finally, an RL policy is used to generate control commands based on the quadrotor's I. Quadrotors are extensively used in a variety of applications, including rescue operations, fire and electricity inspection, and package delivery.

History eXplanation based on Predicates (HXP), studies the behavior of a Reinforcement Learning (RL) agent in a sequence of agent's interactions with the environment (a history), through the prism of an arbitrary predicate. To this end, an action importance score is computed for each action in the history. The explanation consists in displaying the most important actions to the user. As the calculation of an action's importance is #W[1]-hard, it is necessary for long histories to approximate the scores, at the expense of their quality. We therefore propose a new HXP method, called Backward-HXP, to provide explanations for these histories without having to approximate scores. Experiments show the ability of B-HXP to summarise long histories.

In this work, we introduce a control framework that combines model-based footstep planning with Reinforcement Learning (RL), leveraging desired footstep patterns derived from the Linear Inverted Pendulum (LIP) dynamics. Utilizing the LIP model, our method forward predicts robot states and determines the desired foot placement given the velocity commands. We then train an RL policy to track the foot placements without following the full reference motions derived from the LIP model. This partial guidance from the physics model allows the RL policy to integrate the predictive capabilities of the physics-informed dynamics and the adaptability characteristics of the RL controller without overfitting the policy to the template model. Our approach is validated on the MIT Humanoid, demonstrating that our policy can achieve stable yet dynamic locomotion for walking and turning. We further validate the adaptability and generalizability of our policy by extending the locomotion task to unseen, uneven terrain. During the hardware deployment, we have achieved forward walking speeds of up to 1.5 m/s on a treadmill and have successfully performed dynamic locomotion maneuvers such as 90-degree and 180-degree turns.

Social robot navigation (SRN) is a relevant problem that involves navigating a pedestrian-rich environment in a socially acceptable manner. It is an essential part of making social robots effective in pedestrian-rich settings. The use cases of such robots could vary from companion robots to warehouse robots to autonomous wheelchairs. In recent years, deep reinforcement learning has been increasingly used in research on social robot navigation. Our work introduces CAMRL (Context-Aware Mamba-based Reinforcement Learning). Mamba is a new deep learning-based State Space Model (SSM) that has achieved results comparable to transformers in sequencing tasks. CAMRL uses Mamba to determine the robot's next action, which maximizes the value of the next state predicted by the neural network, enabling the robot to navigate effectively based on the rewards assigned. We evaluate CAMRL alongside existing solutions (CADRL, LSTM-RL, SARL) using a rigorous testing dataset which involves a variety of densities and environment behaviors based on ORCA and SFM, thus, demonstrating that CAMRL achieves higher success rates, minimizes collisions, and maintains safer distances from pedestrians. This work introduces a new SRN planner, showcasing the potential for deep-state space models for robot navigation.

Cross-modal knowledge distillation (CMKD) refers to the scenario in which a learning framework must handle training and test data that exhibit a modality mismatch, more precisely, training and test data do not cover the same set of data modalities. Traditional approaches for CMKD are based on a teacher/student paradigm where a teacher is trained on multi-modal data with the aim to successively distill knowledge from a multi-modal teacher to a single-modal student. Despite the widespread adoption of such paradigm, recent research has highlighted its inherent limitations in the context of cross-modal knowledge transfer.Taking a step beyond the teacher/student paradigm, here we introduce a new framework for cross-modal knowledge distillation, named DisCoM-KD (Disentanglement-learning based Cross-Modal Knowledge Distillation), that explicitly models different types of per-modality information with the aim to transfer knowledge from multi-modal data to a single-modal classifier. To this end, DisCoM-KD effectively combines disentanglement representation learning with adversarial domain adaptation to simultaneously extract, foreach modality, domain-invariant, domain-informative and domain-irrelevant features according to a specific downstream task. Unlike the traditional teacher/student paradigm, our framework simultaneously learns all single-modal classifiers, eliminating the need to learn each student model separately as well as the teacher classifier. We evaluated DisCoM-KD on three standard multi-modal benchmarks and compared its behaviourwith recent SOTA knowledge distillation frameworks. The findings clearly demonstrate the effectiveness of DisCoM-KD over competitors considering mismatch scenarios involving both overlapping and non-overlapping modalities. These results offer insights to reconsider the traditional paradigm for distilling information from multi-modal data to single-modal neural networks.

We consider reinforcement learning (RL) methods for finding optimal policies in linear quadratic (LQ) mean field control (MFC) problems over an infinite horizon in continuous time, with common noise and entropy regularization. We study policy gradient (PG) learning and first demonstrate convergence in a model-based setting by establishing a suitable gradient domination condition.Next, our main contribution is a comprehensive error analysis, where we prove the global linear convergence and sample complexity of the PG algorithm with two-point gradient estimates in a model-free setting with unknown parameters. In this setting, the parameterized optimal policies are learned from samples of the states and population distribution.Finally, we provide numerical evidence supporting the convergence of our implemented algorithms.

Conventional uncertainty-aware temporal difference (TD) learning methods often rely on simplistic assumptions, typically including a zero-mean Gaussian distribution for TD errors. Such oversimplification can lead to inaccurate error representations and compromised uncertainty estimation. In this paper, we introduce a novel framework for generalized Gaussian error modeling in deep reinforcement learning, applicable to both discrete and continuous control settings. Our framework enhances the flexibility of error distribution modeling by incorporating higher-order moments, particularly kurtosis, thereby improving the estimation and mitigation of data-dependent noise, i.e., aleatoric uncertainty. We examine the influence of the shape parameter of the generalized Gaussian distribution (GGD) on aleatoric uncertainty and provide a closed-form expression that demonstrates an inverse relationship between uncertainty and the shape parameter. Additionally, we propose a theoretically grounded weighting scheme to fully leverage the GGD. To address epistemic uncertainty, we enhance the batch inverse variance weighting by incorporating bias reduction and kurtosis considerations, resulting in improved robustness. Extensive experimental evaluations using policy gradient algorithms demonstrate the consistent efficacy of our method, showcasing significant performance improvements.

Generalization is a pivotal challenge for agents following natural language instructions. To approach this goal, we leverage a vision-language model (VLM) for visual grounding and transfer its vision-language knowledge into reinforcement learning (RL) for object-centric tasks, which makes the agent capable of zero-shot generalization to unseen objects and instructions. By visual grounding, we obtain an object-grounded confidence map for the target object indicated in the instruction. Based on this map, we introduce two routes to transfer VLM knowledge into RL. Firstly, we propose an object-grounded intrinsic reward function derived from the confidence map to more effectively guide the agent towards the target object. Secondly, the confidence map offers a more unified, accessible task representation for the agent's policy, compared to language embeddings. This enables the agent to process unseen objects and instructions through comprehensible visual confidence maps, facilitating zero-shot object-level generalization. Single-task experiments prove that our intrinsic reward significantly improves performance on challenging skill learning. In multi-task experiments, through testing on tasks beyond the training set, we show that the agent, when provided with the confidence map as the task representation, possesses better generalization capabilities than language-based conditioning. The code is available at https://github.com/PKU-RL/COPL.

The thermal system of battery electric vehicles demands advanced control. Its thermal management needs to effectively control active components across varying operating conditions. While robust control function parametrization is required, current methodologies show significant drawbacks. They consume considerable time, human effort, and extensive real-world testing. Consequently, there is a need for innovative and intelligent solutions that are capable of autonomously parametrizing embedded controllers. Addressing this issue, our paper introduces a learning-based tuning approach. We propose a methodology that benefits from automated scenario generation for increased robustness across vehicle usage scenarios. Our deep reinforcement learning agent processes the tuning task context and incorporates an image-based interpretation of embedded parameter sets. We demonstrate its applicability to a valve controller parametrization task and verify it in real-world vehicle testing. The results highlight the competitive performance to baseline methods. This novel approach contributes to the shift towards virtual development of thermal management functions, with promising potential of large-scale parameter tuning in the automotive industry.

Policy constraint methods in offline reinforcement learning employ additional regularization techniques to constrain the discrepancy between the learned policy and the offline dataset. However, these methods tend to result in overly conservative policies that resemble the behavior policy, thus limiting their performance. We investigate this limitation and attribute it to the static nature of traditional constraints. In this paper, we propose a novel dynamic policy constraint that restricts the learned policy on the samples generated by the exponential moving average of previously learned policies. By integrating this self-constraint mechanism into off-policy methods, our method facilitates the learning of non-conservative policies while avoiding policy collapse in the offline setting. Theoretical results show that our approach results in a nearly monotonically improved reference policy. Extensive experiments on the D4RL MuJoCo domain demonstrate that our proposed method achieves state-of-the-art performance among the policy constraint methods.