ICRA 2023 论文列表

The well-known quote from George Box states that: “All models are wrong, but some are useful.” To develop more useful models, we quantify the inaccuracy with which a given model represents a system of interest, so that we may leverage this quantity to facilitate controller synthesis and verification. Specifically, we develop a procedure that identifies a sim2real gap that holds with a minimum probability. Augmenting the nominal model with our identified sim2real gap produces an uncertain model which we prove is an accurate representor of system behavior. We leverage this uncertain model to synthesize and verify a controller in simulation using a probabilistic verification approach. This pipeline produces controllers with an arbitrarily high probability of realizing desired safe behavior on system hardware without requiring hardware testing except for those required for sim2real gap identification. We also showcase our procedure working on two hardware platforms - the Robotarium and a quadruped.

In this work, we focus on the problem of safe policy transfer in reinforcement learning: we seek to leverage existing policies when learning a new task with specified constraints. This problem is important for safety-critical applications where interactions are costly and unconstrained exploration can lead to undesirable or dangerous outcomes, e.g., with physical robots that interact with humans. We propose a Constrained Markov Decision Process (CMDP) formulation that simultaneously enables the transfer of policies and adherence to safety constraints. Our formulation cleanly separates task goals from safety considerations and permits the specification of a wide variety of constraints. Our approach relies on a novel extension of generalized policy improvement to constrained settings via a Lagrangian formulation. We devise a dual optimization algorithm that estimates the optimal dual variable of a target task, thus enabling safe transfer of policies derived from successor features learned on source tasks. Our experiments in simulated domains show that our approach is effective; it visits unsafe states less frequently and outperforms alternative state-of-the-art methods when taking safety constraints into account.

Programming manipulation behaviors can become increasingly difficult with a growing number and complexity of manipulation tasks, particularly in a dynamic and unstructured environment. Recent progress in unsupervised skill discovery algorithms has shown great promise in learning an extensive collection of behaviors without extrinsic supervision. On the other hand, safety is one of the most critical factors for real- world robot applications. As skill discovery methods typically encourage exploratory and dynamic behaviors, it can often be the case that a large portion of learned skills remain too dangerous and unsafe. In this paper, we introduce the novel problem of Safety-Aware Skill Discovery, which aims to learn, in a task-agnostic fashion, a repertoire of reusable skills that are inherently safe to be composed for solving downstream tasks. We present a computationally tractable algorithm that learns a latent-conditioned skill policy that maximizes intrinsic rewards regularized with a safety-critic that can model any user-defined safety constraints. Using the pretrained safe skill repertoire, hierarchical reinforcement learning can solve multiple downstream tasks without the need for explicit consideration of safety during training and testing. We evaluate our algorithm on a collection of force-controlled robotic manipulation tasks in simulation and show promising downstream task performance while satisfying safety constraints.

We propose a novel class of risk-aware control barrier functions (RA-CBFs) for the control of stochastic safety-critical systems. Leveraging a result from the stochastic level-crossing literature, we deviate from the martingale theory that is currently used in stochastic CBF techniques and prove that a RA-CBF based control synthesis confers a tighter upper bound on the probability of the system becoming unsafe within a finite time interval than existing approaches. We highlight the advantages of our proposed approach over the state-of-the-art via a comparative study on an mobile-robot example, and further demonstrate its viability on an autonomous vehicle highway merging problem in dense traffic.

In this paper, we present an approach for quantifying the propagated uncertainty of robot systems in an online and data-driven manner. Especially in Human-Robot Collaboration, keeping track of the safety compliance during run time is essential: Misclassifying dangerous situations as safe might result in severe accidents. According to official regulations (e.g., ISO standards), safety in industrial robot applications depends on critical parameters, such as the distance and relative velocity between humans and robots. However, safety can only be assured given a measure for the reliability of these parameters. While different risk detection and mitigation approaches exist in literature, a measure that can be used to evaluate safety limits online, and succinctly implies whether a situation is safe or dangerous, is missing to date. Motivated by this, we introduce a generalizable method for calculating the propagated measurement uncertainty of arbitrary parameters, that captures the accumulated uncertainty originating from sensory devices and environmental disturbances of the system. To show that our approach delivers correct results, we perform validation experiments in simulation. In addition, we employ our method in two real-world settings and demonstrate how quantifying the propagated uncertainty of critical parameters facilitates assessing safety online in Human-Robot Collaboration.

Learning-based trajectory planners in robotics have attracted growing interest given their ability to plan for complex tasks. These planners are typically trained in simulation under nominal conditions before being implemented on real robots. However, in real settings, the presence of motion and sensing uncertainties causes the robot to deviate from planned reference trajectories potentially leading to unsafe outcomes such as collisions. In this paper we present a reachability analysis to predict such deviations and to evaluate robot safety along reference trajectories. We then use the reachability analysis to safeguard a learning-based planner. Finally, we demonstrate the applicability of our safeguarding algorithm for learning-based planners via multiple simulations and real robot experiments.

To achieve autonomy in unknown and unstruc-tured environments, we propose a method for semantic-based planning under perceptual uncertainty. This capability is cru-cial for safe and efficient robot navigation in environment with mobility-stressing elements that require terrain-specific locomotion policies. We propose the Semantic Belief Graph (SBG), a geometric- and semantic-based representation of a robot's probabilistic roadmap in the environment. The SBG nodes comprise of the robot geometric state and the semantic-knowledge of the terrains in the environment. The SBG edges represent local semantic-based controllers that drive the robot between the nodes or invoke an information gathering action to reduce semantic belief uncertainty. We formulate a semantic-based planning problem on SBG that produces a policy for the robot to safely navigate to the target location with min-imal traversal time. We analyze our method in simulation and present real-world results with a legged robotic platform navigating multi-level outdoor environments.

The motion planning problem for multiple unstop-pable agents is of interest in many robotics applications, for example, autonomous traffic management for multiple fixed-wing aircraft. Unfortunately, many of the existing algorithms cannot provide safety for such agents, because they require the agents to be able to brake to a complete stop for safety and feasibility insurance. In this paper, we present a distributed multi-agent motion planner that guarantees collision avoidance and persistent feasibility, which can be applied to a team of homogeneous mobile vehicles that cannot stop. The planner is built on top of the idea that a collision-free trajectory in form of a loop can safely accommodate multiple unstoppable agents, while avoiding collisions among them and static obstacles. At every time step, in a distributed manner, the agents generate trajectory-manipulating actions that preserve the loop structure. Then, a deconfliction process selects a conflict-free subset of the generated actions, which are applied at the next time step. Through simulation using an unstoppable Dubins car model, we show that the proposed motion planner is able to provide persistent safety guarantees for such agents in obstacle-cluttered space in real-time.

Industrial insertion tasks are often performed repetitively with parts that are subject to tight tolerances and prone to breakage. Learning an industrial insertion policy in real is challenging as the collision between the parts and the environment can cause slippage or breakage of the part. In this paper, we present a safe self-supervised method to learn a visuo-tactile insertion policy that is robust to grasp pose variations. The method reduces human input and collisions between the part and the receptacle. The method divides the insertion task into two phases. In the first align phase, a tactile-based grasp pose estimation model is learned to align the insertion part with the receptacle. In the second insert phase, a vision-based policy is learned to guide the part into the receptacle. The robot uses force-torque sensing to achieve a safe self-supervised data collection pipeline. Physical experiments on the USB insertion task from the NIST Assembly Taskboard suggest that the resulting policies can achieve 45/45 insertion successes on 45 different initial grasp poses, improving on two baselines: (1) a behavior cloning agent trained on 50 human insertion demonstrations (1/45) and (2) an online RL policy (TD3) trained in real (0/45).

Safety is a fundamental property for the real-world deployment of robotic platforms. Any control policy should avoid dangerous actions that could harm the environment, humans, or the robot itself. In reinforcement learning (RL), safety is crucial when exploring a new environment to learn a new skill. This paper introduces a new formulation of safe exploration for robotic RL in the tangent space of the constraint manifold that effectively transforms the action space of the RL agent for always respecting safety constraints locally. We show how to apply this approach to a wide range of robotic platforms and how to define safety constraints that represent dynamic articulated objects like humans in the context of robotic RL. Our proposed approach achieves state-of-the-art performance in simulated high-dimensional and dynamic tasks while avoiding collisions with the environment. We show safe real-world deployment of our learned controller on a \text{TIAGo}++\text{TIAGo}++ robot, achieving remarkable performance in manipulation and human-robot interaction tasks.

The goal of autonomous vehicles is to navigate public roads safely and comfortably. To enforce safety, traditional planning approaches rely on handcrafted rules to generate trajectories. Machine learning-based systems, on the other hand, scale with data and are able to learn more complex behaviors. However, they often ignore that agents and self-driving vehicle trajectory distributions can be leveraged to improve safety. In this paper, we propose modeling a distribution over multiple future trajectories for both the self-driving vehicle and other road agents, using a unified neural network architecture for prediction and planning. During inference, we select the planning trajectory that minimizes a cost taking into account safety and the predicted probabilities. Our approach does not depend on any rule-based planners for trajectory generation or optimization, improves with more training data and is simple to implement. We extensively evaluate our method through a realistic simulator and show that the predicted trajectory distribution corresponds to different driving profiles. We also successfully deploy it on a self-driving vehicle on urban public roads, confirming that it drives safely without compromising comfort. The code for training and testing our model on a public prediction dataset and the video of the road test are available at https://woven.mobi/safepathnet.

Command and control of an aerial swarm is a complex task. This task increases in difficulty when the flight volume is restricted and the swarm and operator inhabit the same workspace. This work presents a novel method for interacting with and controlling a swarm of quadrotors in a confined space. EMG-based gesture control is used to control the position, orientation, and density of the swarm. Inter-agent as well as agent-operator collisions are prevented through a velocity controller based on a distance-based potential function. State feedback is relayed to the operator via a vibrotactile haptic vest. This cobot human swarm interface prioritizes operator safety while reducing the cognitive load during control of a cobot swarm. This work demonstrates that an operator can safely and intuitively control a swarm of aerial robots in the same workspace.

We propose a policy search approach to learn controllers from specifications given as Signal Temporal Logic (STL) formulae. The system model, which is unknown but assumed to be an affine control system, is learned together with the control policy. The model is implemented as two feedforward neural networks (FNNs) - one for the drift, and one for the control directions. To capture the history dependency of STL specifications, we use a recurrent neural network (RNN) to implement the control policy. In contrast to prevalent model-free methods, the learning approach proposed here takes advantage of the learned model and is more efficient. We use control barrier functions (CBFs) with the learned model to improve the safety of the system. We validate our algorithm via simulations and experiments. The results show that our approach can satisfy the given specification within very few system runs, and can be used for on-line control.

Safe motion control in unknown environments is one of the challenging tasks in robotics, such as autonomous navigation. Control Barrier Function (CBF), as a strong math-ematical tool, has been widely used in many safety-critical systems to satisfy safety requirements. However, there are only a handful of recent studies on safety controllers with perception inputs. Common assumptions in most of the works are that the CBF is already known and obstacles have predefined shapes. In this work, we introduce a novel Vision-based Control Barrier Function (V-CBF), which enables generalization to new environments and obstacles of arbitrary shapes. We then derive CBF safety conditions over RGB-D space and relate those to actual robot control inputs. To train the CBF function, we introduce a method to generate ground truth with desired properties complying with CBF and a method to generate part of the CBF as an image-to-image translation problem. We finally demonstrate the efficacy of V-CBF on the safe control of an autonomous car in CARLA simulator.

Safe path planning is critical for bipedal robots to operate in safety-critical environments. Common path planning algorithms, such as RRT or RRT*, typically use geometric or kinematic collision check algorithms to ensure collision-free paths toward the target position. However, such approaches may generate non-smooth paths that do not comply with the dynamics constraints of walking robots. It has been shown that the control barrier function (CBF) can be integrated with RRT/RRT*to synthesize dynamically feasible collision-free paths. Yet, existing work has been limited to simple circular or elliptical shape obstacles due to the challenging nature of constructing appropriate barrier functions to represent irregularly shaped obstacles. In this paper, we present a CBF-based RRT* algorithm for bipedal robots to generate a collision-free path through space with multiple polynomial-shaped obstacles. In particular, we used logistic regression to construct polynomial barrier functions from a grid map of the environment to represent irregularly shaped obstacles. Moreover, we developed a multi-step CBF steering controller to ensure the efficiency of free space exploration. The proposed approach was first validated in simulation for a differential drive model, and then experimentally evaluated with a 3D humanoid robot, Digit, in a lab setting with randomly placed obstacles.

In this paper, we greatly expand the capability of robots to perform the follow-ahead task and variations of this task through development of a neural network model to predict future human motion from an observed human motion history. We propose a non-autoregressive transformer architecture to leverage its parallel nature for easier training and fast, accurate predictions at test time. The proposed architecture divides human motion prediction into two parts: 1) the human trajectory, which is the 3D positions of the hip joint over time, and 2) the human pose which is the 3D positions of all other joints over time with respect to a fixed hip joint. We propose to make the two predictions simultaneously, as the shared representation can improve the model performance. Therefore, the model consists of two sets of encoders and decoders. First, a multi-head attention module applied to encoder outputs improves human trajectory. Second, another multi-head self-attention module applied to encoder outputs concatenated with decoder outputs facilitates the learning of temporal dependencies. Our model is well-suited for robotic applications in terms of test accuracy and speed, and compares favorably with respect to state-of-the-art methods. We demonstrate the real-world applicability of our work via the Robot Follow-Ahead task, a challenging yet practical case study for our proposed model. The human motion predicted by our model enables the robot follow-ahead in scenarios that require taking detailed human motion into account such as sit-to-stand, stand-to-sit. It also enables simple control policies to trivially generalize to many different variations of human following, such as follow-beside. Our code and data are available at the following Github page: https://github.com/mmahdavian/STPOTR

Vibration perception is essential for robotic sensing and dynamic control. Nevertheless, due to the rigorous demand for sensor conformability and stretchability, enabling soft robots with proprioceptive vibration sensing remains challenging. This paper proposes a novel liquid metal-based stretchable e-skin via a kirigami-inspired design to enable soft robot proprioceptive vibration sensing. The e-skin is fabricated into 0.1mm ultrathin thickness, ensuring its negligible influence on the overall stiffness of the soft robot. Moreover, the working mechanism of the e-skin is based on the ubiquitous triboelectrification effect, which transduces mechanical stimuli without external power supply. To demonstrate the practicability of the e-skin, we built a soft gripper consisting of three soft robotic fingers with proprioceptive vibration sensing. Our experiment shows that the gripper can accurately distinguish the grain category (six grains with the same mass, 99.9% accuracy) and the packaging quality (100% accuracy) by simply shaking the gripped bottle. In summary, a soft robotic proprioceptive vibration sensing solution is proposed; it helps soft robots to have a more comprehensive awareness of their self-state and may inspire further research on soft robotics.

Self-supervised depth estimation draws a lot of attention recently as it can promote the 3D sensing capa-bilities of self-driving vehicles. However, it intrinsically relies upon the photometric consistency assumption, which hardly holds during nighttime. Although various supervised night-time image enhancement methods have been proposed, their generalization performance in challenging driving scenarios is not satisfactory. To this end, we propose the first method that jointly learns a nighttime image enhancer and a depth estimator, without using ground truth for either task. Our method tightly entangles two self-supervised tasks using a newly proposed uncertain pixel masking strategy. This strategy originates from the observation that nighttime images not only suffer from underexposed regions but also from overexposed regions. By fitting a bridge-shaped curve to the illumination map distribution, both regions are suppressed and two tasks are bridged naturally. We benchmark the method on two established datasets: nuScenes and RobotCar and demonstrate state-of-the-art performance on both of them. Detailed ablations also reveal the mechanism of our proposal. Last but not least, to mitigate the problem of sparse ground truth of existing datasets, we provide a new photo-realistically enhanced nighttime dataset based upon CARLA. It brings meaningful new challenges to the community. Codes, data, and models are available at https://github.com/ucaszyp/STEPS.

In this work, we present a novel global descriptor termed stable triangle descriptor (STD) for 3D place recognition. For a triangle, its shape is uniquely determined by the length of the sides or included angles. Moreover, the shape of triangles is completely invariant to rigid transformations. Based on this property, we first design an algorithm to efficiently extract local key points from the 3D point cloud and encode these key points into triangular descriptors. Then, place recognition is achieved by matching the side lengths (and some other information) of the descriptors between point clouds. The point correspondence obtained from the descriptor matching pair can be further used in geometric verification, which greatly improves the accuracy of place recognition. In our experiments, we extensively compare our proposed system against other state-of-the-art systems (i.e., M2DP, Scan Context) on public datasets (i.e., KITTI, NCLT, and Complex-Urban) and our self-collected dataset (with a non-repetitive scanning solid-state LiDAR). All the quantitative results show that STD has stronger adaptability and a great improvement in precision over its counterparts. To share our findings and make contributions to the community, we open source our code on our GitHub: github.com/hku-mars/STD.