Recent advances in control and planning allow for seamless physical human–robot interaction (pHRI). At the same time, novel challenges appear in orchestrating intelligent decision-making and ensuring safe control of robots. Particularly in scenarios involving unforeseen or unintended collisions, robots face the imperative of reacting judiciously to avert potential risks to humans, other robots, obstacles, or themselves. At the same time, they need to maintain focus on their primary task or be able to safely resume it. Collision detection and identification algorithms are now well established in industry, yet complex collision reflexes have not transitioned into industrial applications beyond basic stopping reactions. Despite the introduction of numerous advanced high-performance reflex controllers over the past decades, their real-world adoption has remained a challenge. This work establishes a systematic framework to address that gap. For this, thereflex control problemis defined,reflex behaviorsare systematically classified and categorized, and relevantsafety datais acquired followingexisting international standards. We argue that this foundational step is crucial for improving the safety and capabilities of robots in both complex industrial and domestic environments. We validate our approach within the system class of articulated manipulators through a state-of-the-art cooperative pick-and-place task, providing a blueprint for future implementations for other robot classes.

Future robots operating in fast-changing anthropomorphic environments need to be reactive, safe, flexible, and intuitively use both arms (comparable to humans) to handle task-space constrained manipulation scenarios. Furthermore, dynamic environments pose additional challenges for motion planning due to a continual requirement for validation and refinement of plans. This work addresses the issues with vector-field-based motion generation strategies, which are often prone to local-minima problems. We aim to bridge the gap between reactive solutions, global planning, and constrained cooperative (two-arm) manipulation in partially known surroundings. To this end, we introduce novel planning and real-time control strategies leveraging the geometry of the task space that are inherently coupled for seamless operation in dynamic scenarios. Our integrated multiagent global planning and control scheme explores controllable sets in the previously introducedcooperative dual task spaceand flexibly controls them by exploiting the redundancy of the high degree-of-freedom (DOF) system. The planning and control framework is extensively validated in complex, cluttered, and nonstationary simulation scenarios where our framework is able to complete constrained tasks in a reliable manner, whereas existing solutions fail. We also perform additional real-world experiments with a two-armed 14 DOF torque-controlled KoBo robot. Our rigorous simulation studies and real-world experiments reinforce the claim that the framework is able to run robustly within the inner loop of modern collaborative robots with vision feedback.

Knowledge of the robot's load inertial parameters is indispensable for accurate and safe operation, especially in collaborative robotics. However, an intuitive method for online inertial payload identification, usable while the robot is executing another online generated task, is still lacking. In this work, we propose an online payload identification approach based on the momentum observer using proprioceptive sensors of tactile robots and a novel filter design of kinematic measure-ments. Furthermore, we introduce a novel calibration scheme, that allows circumventing constraints of current calibration methods for payload identification. Specifically, the requirement of performing exactly the same motion for calibration as well as for the identification process is released. This is achieved by introducing an average virtual calibration object that improves the robot model for the identification process. In experiments with a Franka Emika Panda robot, it is shown that the proposed methods surpass common methods in terms of identification error. Especially, the novel calibration approach shows high robustness against temporal and spatial misalignment of the motions.

In order to transform a robot into an intelligent machine it needs to be enabled to react to unforeseen events (most importantly collisions) during task execution and have a plan on how to continue the task afterwards. This requires a flexible operational framework that allows to define adaptive reactions and interactions with the motion generation and task planning stage. Within this work we first reason about the choices the robot has for reactions to unforeseen events such as collisions with respect to safety of humans in the workspace, the robot itself and the environment as well as the successful task execution. We further present a flexible reflex engine together with a concept of integration into the motion generation and control work flow. The reflex engine and it's reflex maneuvers are a combination of state machines and decision trees that take into account the state of the robot and the world. It is capable of choosing safe reactions and can differentiate between different levels of contact severity and according reaction sets. Several reflex maneuvers are evaluated towards safety performance criteria in real robot experiments using an ISO/TS 15066 conform measurement device. Some of the tested reflexes are furthermore integrated into an implementation of the proposed approach for a simple real world example task where the robot needs to pickup a container and dispose it's content into a bin.

Similar to human work, robotic tasks sometimes require two hands to be accomplished. This requires coordinated motion planning and control. While fulfilling the task in a coordinated manner is already a big challenge, the task at hand becomes even harder when obstacles are introduced in the environment that need to be avoided. Furthermore in the case of dynamic environments, contacts cannot be avoided all the time, even with robust planning. In addition to geometric constraints, bimanual systems need to be able to detect and react to contacts during task execution. To this aim, we integrate a vector-field based planning scheme, that is able to avoid obstacles, with contact detection and reactive control methods based on contact wrench estimation such as admittance control. We also fuse the real contact forces into the planner directly together with the circular repulsive fields. The resulting planner-controller combination is capable of obstacle avoidance planning as well as reaction control in the case of unforeseen contacts that can also be used in situations where the manipulation needs to be guided by the environment such as landing control in only roughly known environments. We evaluate our approach on the torque-controlled Kobo bimanual set-up and also perform rigorous simulation studies.

Humanoid robots in unknown environments need to be able to quickly react to contacts in order to ensure safety of humans and their own hardware. For showing useful reactions to contacts, the robot needs information about possibly multiple contacts such as their respective contact locations and wrenches. In this paper, we introduce our algorithm rm-Code, a real-time multi-contact detection, isolation and identification algorithm for tree-structured floating base robots based on generalized external forces and (optional) external wrenches measured by force/torque sensors within the kinematic chain. Those entities have been deduced in the literature using proprioceptive sensing only. Furthermore, the algorithm is fast enough for online computation. To the best of our knowledge, this is the first algorithm combining all of these properties. Rm-Code is quantitatively evaluated in simulation. The results show that our solution is able to accurately solve the problem when fed with perfect input data. In a second step, possible sources of error in the presence of noisy input data are analyzed. It is concluded that purely proprioception based contact isolation and identification in the multi-contact case has certain limitations under realistic conditions. However, these limitations could be overcome easily by integrating simple link contact detection, e.g. bumpers or other similarly simple means.

High-performance collision handling, which is divided into the five phases detection, isolation, estimation, classification and reaction, is a fundamental robot capability for safe and sensitive operation/interaction in unknown environments. For complex humanoid robots collision handling is obviously significantly more complex than for classical static manipulators. In particular, the robot stability during the collision reaction phase has to be carefully designed and relies on high fidelity contact information that is generated during the first three phases. In this paper, a unified realtime algorithm is presented for determining unknown contact forces and contact locations for humanoid robots based on proprioceptive sensing only, i.e. joint position, velocity and torque, as well as force/torque sensing along the structure. The proposed scheme is based on nonlinear model-based momentum observers that are able to recover the unknown contact forces and the respective locations. The dynamic loads acting on internal force/torque sensors are also corrected based on a novel nonlinear compensator. The theoretical capabilities of the presented methods are evaluated in simulation with the Atlas robot. In summary, we propose a full solution to the problem of collision detection, collision isolation and collision identification for the general class of humanoid robots.

Soft robotics methods such as impedance control and reflexive collision handling have proven to be a valuable tool to robots acting in partially unknown and potentially unstructured environments. Mainly, the schemes were developed with focus on classical electromechanically driven, torque controlled robots. There, joint friction, mostly coming from high gearing, is typically decoupled from link-side control via suitable rigid or elastic joint torque feedback. Extending and applying these algorithms to stiff hydraulically actuated robots poses problems regarding the strong influence of friction on joint torque estimation from pressure sensing, i.e. link-side friction is typically significantly higher than in electromechanical soft robots. In order to improve the performance of such systems, we apply state-of-the-art fault detection and estimation methods together with observer-based disturbance compensation control to the humanoid robot Atlas. With this it is possible to achieve higher tracking accuracy despite facing significant modeling errors. Compliant end-effector behavior can also be ensured by including an additional force/torque sensor into the generalized momentum-based disturbance observer algorithm from [1].