Robotics (cs.RO)

Abstract: In recent years, Deep Reinforcement Learning (DRL) has emerged as a promising method for robot collision avoidance. However, such DRL models often come with limitations, such as adapting effectively to structured environments containing various pedestrians. In order to solve this difficulty, previous research has attempted a few approaches, including training an end-to-end solution by integrating a waypoint planner with DRL and developing a multimodal solution to mitigate the drawbacks of the DRL model. However, these approaches have encountered several issues, including slow training times, scalability challenges, and poor coordination among different models. To address these challenges, this paper introduces a novel approach called evolutionary curriculum training to tackle these challenges. The primary goal of evolutionary curriculum training is to evaluate the collision avoidance model's competency in various scenarios and create curricula to enhance its insufficient skills. The paper introduces an innovative evaluation technique to assess the DRL model's performance in navigating structured maps and avoiding dynamic obstacles. Additionally, an evolutionary training environment generates all the curriculum to improve the DRL model's inadequate skills tested in the previous evaluation. We benchmark the performance of our model across five structured environments to validate the hypothesis that this evolutionary training environment leads to a higher success rate and a lower average number of collisions. Further details and results at our project website.

Abstract: This work introduces a flexible architecture for real-time occupancy forecasting. In contrast to existing, more computationally expensive architectures, the proposed model exploits recursive latent state estimation, using learned transformer-based prediction and update modules. This allows for highly efficient real-time inference on an embedded system (profiled on an Nvidia Xavier AGX), and the inclusion of a broad set of information from a diverse set of sensors. The architecture is able to process sparse and occluded observations of agent positions and scene context as this is made available, and does not require motion tracklet inputs. \networkName{} accomplishes this by encoding the scene into a latent state that evolves in time with self-attention and is updated with contextual information such as traffic signals, road topology or agent detections using cross-attention. Occupancy predictions are made by sparsely querying positions of interest as opposed to generating a fixed size raster image, which allows for variable resolution occupancy prediction or local querying by downstream trajectory optimisation algorithms, saving computational effort.

Abstract: Wheeled robot navigation has been widely used in urban environments, but little research has been conducted on its navigation in wild vegetation. External sensors (LiDAR, camera etc.) are often used to construct point cloud map of the surrounding environment, however, the supporting rigid ground used for travelling cannot be detected due to the occlusion of vegetation. This often causes unsafe or not smooth path during planning process. To address the drawback, we propose the PE-RRT* algorithm, which effectively combines a novel support plane estimation method and sampling algorithm to generate real-time feasible and safe path in vegetation environments. In order to accurately estimate the support plane, we combine external perception and proprioception, and use Multivariate Gaussian Processe Regression (MV-GPR) to estimate the terrain at the sampling nodes. We build a physical experimental platform and conduct experiments in different outdoor environments. Experimental results show that our method has high safety, robustness and generalization.

Abstract: Delayed Markov decision processes fulfill the Markov property by augmenting the state space of agents with a finite time window of recently committed actions. In reliance with these state augmentations, delay-resolved reinforcement learning algorithms train policies to learn optimal interactions with environments featured with observation or action delays. Although such methods can directly be trained on the real robots, due to sample inefficiency, limited resources or safety constraints, a common approach is to transfer models trained in simulation to the physical robot. However, robotic simulations rely on approximated models of the physical systems, which hinders the sim2real transfer. In this work, we consider various uncertainties in the modelling of the robot's dynamics as unknown intrinsic disturbances applied on the system input. We introduce a disturbance-augmented Markov decision process in delayed settings as a novel representation to incorporate disturbance estimation in training on-policy reinforcement learning algorithms. The proposed method is validated across several metrics on learning a robotic reaching task and compared with disturbance-unaware baselines. The results show that the disturbance-augmented models can achieve higher stabilization and robustness in the control response, which in turn improves the prospects of successful sim2real transfer.

Abstract: This paper presents a Predictive Maneuver Planning with Deep Reinforcement Learning (PMP-DRL) model for maneuver planning. Traditional rule-based maneuver planning approaches often have to improve their abilities to handle the variabilities of real-world driving scenarios. By learning from its experience, a Reinforcement Learning (RL)-based driving agent can adapt to changing driving conditions and improve its performance over time. Our proposed approach combines a predictive model and an RL agent to plan for comfortable and safe maneuvers. The predictive model is trained using historical driving data to predict the future positions of other surrounding vehicles. The surrounding vehicles' past and predicted future positions are embedded in context-aware grid maps. At the same time, the RL agent learns to make maneuvers based on this spatio-temporal context information. Performance evaluation of PMP-DRL has been carried out using simulated environments generated from publicly available NGSIM US101 and I80 datasets. The training sequence shows the continuous improvement in the driving experiences. It shows that proposed PMP-DRL can learn the trade-off between safety and comfortability. The decisions generated by the recent imitation learning-based model are compared with the proposed PMP-DRL for unseen scenarios. The results clearly show that PMP-DRL can handle complex real-world scenarios and make better comfortable and safe maneuver decisions than rule-based and imitative models.

Abstract: This paper presents a localization algorithm for autonomous urban vehicles under rain weather conditions. In adverse weather, human drivers anticipate the location of the ego-vehicle based on the control inputs they provide and surrounding road contextual information. Similarly, in our approach for localization in rain weather, we use visual data, along with a global reference path and vehicle motion model for anticipating and better estimating the pose of the ego-vehicle in each frame. The global reference path contains useful road contextual information such as the angle of turn which can be potentially used to improve the localization accuracy especially when sensors are compromised. We experimented on the Oxford Robotcar Dataset and our internal dataset from Singapore to validate our localization algorithm in both clear and rain weather conditions. Our method improves localization accuracy by 50.83% in rain weather and 34.32% in clear weather when compared to baseline algorithms.

Abstract: This paper invented a new rolling control mechanism of two section articulated-wing ornithopter, which is analogues to aileron control in plane, however, similar control mechanism leads to opposite result, indicating the ornithopter supposed to go left now go right instead. This research gives a qualitative dynamics model which explains this new phenomenon. Because of wing folding, the differential rotation of outer-section wing (analogues to aileron in plane, left aileron up and right aileron down make left turn) around pitch axis becomes common mode rotation around yaw axis,leading its rotating torque changing from left-handed rotation (using left-handed as example, right-handed is the same) around roll axis to a common mode force pointing to front-right (northeast, NE) direction from first player's view of the ornithopter.Because most of the flapping movement is in the upper hemisphere from ornithopter's view, the NE force is above on the center of mass of the orthopter, generating a right-handed moment around roll axis. Therefore, the ornithopter supposed to go left now goes right. This phenomenon is a unique and only observed in two section articulated-wing ornithopter by far. Many field tests conducted by authors confirm it is highly repetitive.

Abstract: Despite recent progress improving the efficiency and quality of motion planning, planning collision-free and dynamically-feasible trajectories in partially-mapped environments remains challenging, since constantly replanning as unseen obstacles are revealed during navigation both incurs significant computational expense and can introduce problematic oscillatory behavior. To improve the quality of motion planning in partial maps, this paper develops a framework that augments sampling-based motion planning to leverage a high-level discrete layer and prior solutions to guide motion-tree expansion during replanning, affording both (i) faster planning and (ii) improved solution coherence. Our framework shows significant improvements in runtime and solution distance when compared with other sampling-based motion planners.

Abstract: The prominence of embodied Artificial Intelligence (AI), which empowers robots to navigate, perceive, and engage within virtual environments, has attracted significant attention, owing to the remarkable advancements in computer vision and large language models. Privacy emerges as a pivotal concern within the realm of embodied AI, as the robot access substantial personal information. However, the issue of privacy leakage in embodied AI tasks, particularly in relation to decision-making algorithms, has not received adequate consideration in research. This paper aims to address this gap by proposing an attack on the Deep Q-Learning algorithm, utilizing gradient inversion to reconstruct states, actions, and Q-values. The choice of using gradients for the attack is motivated by the fact that commonly employed federated learning techniques solely utilize gradients computed based on private user data to optimize models, without storing or transmitting the data to public servers. Nevertheless, these gradients contain sufficient information to potentially expose private data. To validate our approach, we conduct experiments on the AI2THOR simulator and evaluate our algorithm on active perception, a prevalent task in embodied AI. The experimental results convincingly demonstrate the effectiveness of our method in successfully recovering all information from the data across all 120 room layouts.

Abstract: Motion forecasting plays a critical role in enabling robots to anticipate future trajectories of surrounding agents and plan accordingly. However, existing forecasting methods often rely on curated datasets that are not faithful to what real-world perception pipelines can provide. In reality, upstream modules that are responsible for detecting and tracking agents, and those that gather road information to build the map, can introduce various errors, including misdetections, tracking errors, and difficulties in being accurate for distant agents and road elements. This paper aims to uncover the challenges of bringing motion forecasting models to this more realistic setting where inputs are provided by perception modules. In particular, we quantify the impacts of the domain gap through extensive evaluation. Furthermore, we design synthetic perturbations to better characterize their consequences, thus providing insights into areas that require improvement in upstream perception modules and guidance toward the development of more robust forecasting methods.