Preference-based reinforcement learning (RL) poses as a recent research direction in robot learning, by allowing humans to teach robots through preferences on pairs of desired behaviours. Nonetheless, to obtain realistic robot policies, an arbitrarily large number of queries is required to be answered by humans. In this work, we approach the sample-efficiency challenge by presenting a technique which synthesizes queries, in a semi-supervised learning perspective. To achieve this, we leverage latent variational autoencoder (VAE) representations of trajectory segments (sequences of state-action pairs). Our approach manages to produce queries which are closely aligned with those labeled by humans, while avoiding excessive uncertainty according to the human preference predictions as determined by reward estimations. Additionally, by introducing variation without deviating from the original human's intents, more robust reward function representations are achieved. We compare our approach to recent state-of-the-art preference-based RL semi-supervised learning techniques. Our experimental findings reveal that we can enhance the generalization of the estimated reward function without requiring additional human intervention. Lastly, to confirm the practical applicability of our approach, we conduct experiments involving actual human users in a simulated social navigation setting. Videos of the experiments can be found at https://sites.google.com/view/rl-sequel

@en

Safety is crucial for autonomous drones to operate close to humans. Besides avoiding unwanted or harmful contact, people should also perceive the drone as safe. Existing safe motion planning approaches for autonomous robots, such as drones, have primarily focused on ensuring physical safety, e.g., by imposing constraints on motion planners. However, studies indicate that ensuring physical safety does not necessarily lead to perceived safety. Prior work in Human-Drone Interaction (HDI) shows that factors such as the drone's speed and distance to the human are important for perceived safety. Building on these works, we propose a parameterized control barrier function (CBF) that constrains the drone's maximum deceleration and minimum distance to the human and update its parameters on people's ratings of perceived safety. We describe an implementation and evaluation of our approach. Results of a withinsubject user study (Ng= 15) show that we can improve perceived safety of a drone by adjusting to people individually.

@en

Rearranging objects is an essential skill for robots. To quickly teach robots new rearrangements tasks, we would like to generate training scenarios from high-level specifications that define the relative placement of objects for the task at hand. Ideally, to guide the robot's learning we also want to be able to rank these scenarios according to their difficulty. Prior work has shown how generating diverse scenario from specifications and providing the robot with easy-to-difficult samples can improve the learning. Yet, existing scenario generation methods typically cannot generate diverse scenarios while controlling their difficulty. We address this challenge by conditioning generative models on spatial logic specifications to generate spatially-structured scenarios that meet the specification and desired difficulty level. Our experiments showed that generative models are more effective and data-efficient than rejection sam-pling and that the spatially-structured scenarios can drastically improve training of downstream tasks by orders of magnitude.

@en

Ensuring safety in real-world robotic systems is often challenging due to unmodeled disturbances and noisy sensors. To account for such stochastic uncertainties, many robotic systems leverage probabilistic state estimators such as Kalman filters to obtain a robot's belief, i.e. a probability distribution over possible states. We propose belief control barrier functions (BCBFs) to enable risk-aware control, leveraging all information provided by state estimators. This allows robots to stay in predefined safety regions with desired confidence under these stochastic uncertainties. BCBFs are general and can be applied to a variety of robots that use extended Kalman filters as state estimator. We demonstrate BCBFs on a quadrotor that is exposed to external disturbances and varying sensing conditions. Our results show improved safety compared to traditional state-based approaches while allowing control frequencies of up to 1 kHz.

@en

In this article, we address the task and safety performance of data-driven model predictive controllers (DD-MPC) for systems with complex dynamics, i.e., temporally or spatially varying dynamics that may also be discontinuous. The three challenges we focus on are the accuracy of learned models, the receding horizon-induced myopic predictions of DD-MPC, and the active encouragement of safety. To learn accurate models for DD-MPC, we cautiously, yet effectively, explore the dynamical system with rapidly exploring random trees (RRT) to collect a uniform distribution of samples in the state-input space and overcome the common distribution shift in model learning. The learned model is further used to construct an RRT tree that estimates how close the model's predictions are to the desired target. This information is used in the cost function of the DD-MPC to minimize the short-sighted effect of its receding horizon nature. To promote safety, we approximate sets of safe states using demonstrations of exclusively safe trajectories, i.e., without unsafe examples, and encourage the controller to generate trajectories close to the sets. As a running example, we use a broken version of an inverted pendulum where the friction abruptly changes in certain regions. Furthermore, we showcase the adaptation of our method to a real-world robotic application with complex dynamics: robotic food-cutting. Our results show that our proposed control framework effectively avoids unsafe states with higher success rates than baseline controllers that employ models from controlled demonstrations and even random actions.@en

Practical implementations of deep reinforcement learning (deep RL) have been challenging due to an amplitude of factors, such as designing reward functions that cover every possible interaction. To address the heavy burden of robot reward engineering, we aim to leverage subjective human preferences gathered in the context of human-robot interaction, while taking advantage of a baseline reward function when available. By considering baseline objectives to be designed beforehand, we are able to narrow down the policy space, solely requesting human attention when their input matters the most. To allow for control over the optimization of different objectives, our approach contemplates a multi-objective setting. We achieve human-compliant policies by sequentially training an optimal policy from a baseline specification and collecting queries on pairs of trajectories. These policies are obtained by training a reward estimator to generate Pareto optimal policies that include human preferred behaviours. Our approach ensures sample efficiency and we conducted a user study to collect real human preferences, which we utilized to obtain a policy on a social navigation environment.@en

Many tasks require robots to manipulate objects while satisfying a complex interplay of spatial and temporal constraints. For instance, a table setting robot first needs to place a mug and then fill it with coffee, while satisfying spatial relations such as forks need to placed left of plates. We propose the spatio-temporal framework SpaTiaL that unifies the specification, monitoring, and planning of object-oriented robotic tasks in a robot-agnostic fashion. SpaTiaL is able to specify diverse spatial relations between objects and temporal task patterns. Our experiments with recorded data, simulations, and real robots demonstrate how SpaTiaL provides real-time monitoring and facilitates online planning. SpaTiaL is open source and easily expandable to new object relations and robotic applications.

@en

In many real-world robotic scenarios, we cannot assume exact knowledge about a robot’s state due to unmodeled dynamics or noisy sensors. Planning in belief space addresses this problem by tightly coupling perception and planning modules to obtain trajectories that take into account the environment’s stochasticity. However, existing works are often limited to tasks such as the classic reach-avoid problem and do not provide risk awareness. We propose a risk-aware planning strategy in belief space that minimizes the risk of violating a given specification and enables a robot to actively gather information about its state. We use Risk Signal Temporal Logic (RiSTL) as a specification language in belief space to express complex spatio-temporal missions including predicates over Gaussian beliefs. We synthesize trajectories for challenging scenarios that cannot be expressed through classical reach-avoid properties and show that risk-aware objectives improve the uncertainty reduction in a robot’s belief.@en

Hidden traffic participants pose a great challenge for autonomous vehicles. Previous methods typically do not use previous obser-vations, leading to over-conservative behavior. In this paper, we present a continuation of our work on reasoning about objects out-side the current sensor view. We aim to demonstrate our recently proposed method on an autonomous platform and evaluate its relia-bility and real-time feasibility when using real sensor data. Showing a significant driving performance increase on a real platform, with-out compromising safety, would be a significant contribution to the field of autonomous driving.

@en

Safe driving requires autonomous vehicles to anticipate potential hidden traffic participants and other unseen objects, such as a cyclist hidden behind a large vehicle, or an object on the road hidden behind a building. Existing methods are usually unable to consider all possible shapes and orientations of such obstacles. They also typically do not reason about observations of hidden obstacles over time, leading to conservative anticipations. We overcome these limitations by (1) modeling possible hidden obstacles as a set of states of a point mass model and (2) sequential reasoning based on reachability analysis and previous observations. Based on (1), our method is safer, since we anticipate obstacles of arbitrary unknown shapes and orientations. In addition, (2) increases the available drivable space when planning trajectories for autonomous vehicles. In our experiments, we demonstrate that our method, at no expense of safety, gives rise to significant reductions in time to traverse various intersection scenarios from the CommonRoad Benchmark Suite.

@en

Reinforcement learning has shown great potential for learning sequential decision-making tasks. Yet, it is difficult to anticipate all possible real-world scenarios during training, causing robots to inevitably fail in the long run. Many of these failures are due to variations in the robot's environment. Usually experts are called to correct the robot's behavior; however, some of these failures do not necessarily require an expert to solve them. In this work, we query non-experts online for help and explore 1) if/how non-experts can provide feedback to the robot after a failure and 2) how the robot can use this feedback to avoid such failures in the future by generating shields that restrict or correct its high-level actions. We demonstrate our approach on common daily scenarios of a simulated kitchen robot. The results indicate that non-experts can indeed understand and repair robot failures. Our generated shields accelerate learning and improve data-efficiency during retraining.

@en

Runtime enforcement refers to the theories, techniques, and tools for enforcing correct behavior with respect to a formal specification of systems at runtime. In this paper, we are interested in techniques for constructing runtime enforcers for the concrete application domain of enforcing safety in AI. We discuss how safety is traditionally handled in the field of AI and how more formal guarantees on the safety of a self-learning agent can be given by integrating a runtime enforcer. We survey a selection of work on such enforcers, where we distinguish between approaches for discrete and continuous action spaces. The purpose of this paper is to foster a better understanding of advantages and limitations of different enforcement techniques, focusing on the specific challenges that arise due to their application in AI. Finally, we present some open challenges and avenues for future work.

@en

Despite the successes of deep reinforcement learning (RL), it is still challenging to obtain safe policies. Formal verification approaches ensure safety at all times, but usually overly restrict the agent's behaviors, since they assume adversarial behavior of the environment. Instead of assuming adversarial behavior, we suggest to focus on perceived safety instead, i.e., policies that avoid undesired behaviors while having a desired level of conservativeness. To obtain policies that are perceived as safe, we propose a shield synthesis framework with two distinct loops: (1) an inner loop that trains policies with a set of actions that is constrained by shields whose conservativeness is parameterized, and (2) an outer loop that presents example rollouts of the policy to humans and collects their feedback to update the parameters of the shields in the inner loop. We demonstrate our approach on a RL benchmark of Lunar landing and a scenario in which a mobile robot navigates around humans. For the latter, we conducted two user studies to obtain policies that were perceived as safe. Our results indicate that our framework converges to policies that are perceived as safe, is robust against noisy feedback, and can query feedback for multiple policies at the same time.

@en

Visual planning approaches have shown great success for decision making tasks with no explicit model of the state space. Learning a suitable representation and constructing a latent space where planning can be performed allows non-experts to setup and plan motions by just providing images. However, learned latent spaces are usually not semantically-interpretable, and thus it is difficult to integrate task constraints. We propose a novel framework to determine whether plans satisfy constraints given demonstrations of policies that satisfy or violate the constraints. The demonstrations are realizations of Linear Temporal Logic formulas which are employed to train Long Short-Term Memory (LSTM) networks directly in the latent space representation. We demonstrate that our architecture enables designers to easily specify, compose and integrate task constraints and achieves high performance in terms of accuracy. Furthermore, this visual planning framework enables human interaction, coping the environment changes that a human worker may involve. We show the flexibility of the method on a box pushing task in a simulated warehouse setting with different task constraints.

@en

Safe motion planning for autonomous vehicles is a challenging task, since the exact future motion of other traffic participant is usually unknown. In this article, we present a verification technique ensuring that autonomous vehicles do not cause collisions by using fail-safe trajectories. Fail-safe trajectories are executed if the intended motion of the autonomous vehicle causes a safety-critical situation. Our verification technique is real-time capable and operates under the premise that intended trajectories are only executed if they have been verified as safe. The benefits of our proposed approach are demonstrated in different scenarios on an actual vehicle. Moreover, we present the first in-depth analysis of our verification technique used in dense urban traffic. Our results indicate that fail-safe motion planning has the potential to drastically reduce accidents while not resulting in overly conservative behaviors of the autonomous vehicle.

@en

In this paper, we address the safety of data-driven control for contact-rich manipulation. We propose to restrict the controller's action space to keep the system in a set of safe states. In the absence of an analytical model, we show how Gaussian Processes (GP) can be used to approximate safe sets. We disable inputs for which the predicted states are likely to be unsafe using the GP. Furthermore, we show how locally designed feedback controllers can be used to improve the execution precision in the presence of modelling errors. We demonstrate the benefits of our method on a pushing task with a variety of dynamics, by using known and unknown surfaces and different object loads. Our results illustrate that the proposed approach significantly improves the performance and safety of the baseline controller.

@en

Adaptive cruise control is one of the most common comfort features of road vehicles. Despite its large market penetration, current systems are not safe in all driving conditions and require supervision by human drivers. While several previous works have proposed solutions for safe adaptive cruise control, none of these works considers comfort, especially in the event of cut-ins. We provide a novel solution that simultaneously meets our specifications and provides comfort in all driving conditions, including cut-ins. This is achieved by an exchangeable nominal controller ensuring comfort, combined with a provably correct fail-safe controller that gradually engages an emergency maneuver - this ensures comfort, since most threats are already cleared before emergency braking is fully activated. As a consequence, one can easily exchange the nominal controller without having to have the overall system safety re-certified. We also provide the first user study into a provably-correct adaptive cruise controller. It shows that even though our approach never causes an accident, passengers rate the performance as good as a state-of-the-art solution that does not ensure safety.

@en