This paper presents a feed forward artificial neural network that identifies the order of the dynamics of a unit step response. The main contribution of this paper is demonstrating that a system trained on only integer order (first and second order) systems can identify fractional order responses with a high degree of accuracy. The details of the design of structure of the neural network, the training method and the training sets, as well as statistics describing the accuracy of the fractional predictions are presented. Also using the neural network to identify fractional dynamics for a large scale networked system from the authors’ prior work is presented as further validation and a demonstration of the applicability of the results. This demonstrates the potential for practicing engineers to use similar machine learning tools trained on "standard" systems with the ability to distinguish when features such as fractional order dynamics are significant and warrant deeper consideration for the design or control of such a system.

Collaborative robots require effective human intention estimation to safely and smoothly work with humans in less structured tasks such as industrial assembly, where human intention continuously changes. We propose the concept of intention tracking and introduce a collaborative robot system that concurrently tracks intentions at hierarchical levels. The high-level intention is tracked to estimate human's interaction pattern and enable robot to (1) avoid collision with human to minimize interruption and (2) assist human to correct failure. The low-level intention estimate provides robot with task-related information. We implement the system on a UR5e robot and demonstrate robust, seamless and ergonomic human-robot collaboration in an ablative pilot study of an assembly use case.

In this article, we present a multi-robot task and motion planning method that, when applied to the rearrangement of objects by manipulators, results in solution times up to three orders of magnitude faster than the existing methods and successfully plans for problems with up to 20 objects, more than three times as many objects as comparable methods. We achieve this improvement by decomposing the planning space to consider manipulators alone, objects, and manipulators holding objects. We represent this decomposition with a hypergraph where vertices are decomposed elements of the planning spaces and hyperarcs are transitions between elements. The existing methods use graph-based representations where vertices are full composite spaces and edges are transitions between these. Using the hypergraph reduces the representation size of the planning space for multimanipulator object rearrangement, the number of hypergraph vertices scales linearly with the number of either robots or objects, while the number of hyperarcs scales quadratically with the number of robots and linearly with the number of objects. In contrast, the number of vertices and edges in graph-based representations scales exponentially in the number of robots and objects. We show that similar gains can be achieved for other multi-robot task and motion planning problems.

This paper presents a new method for trajectory planning and control of a 2R planar horizontal robotic arm with only the first joint actuated, which is known to be controllable but not asymptotically stabilizable with smooth feedback controllers. The main idea is to first determine an admissible trajectory that satisfies both the constraints of system dynamics and boundary value conditions, and then design a feedforward controller for the system. For complex and highly nonlinear systems, an optimization method is applied to determine the trajectory. Moreover, a time-scaling method can be used to control the magnitude of the inputs and also make it easier to find a solution in the optimization procedure. To further improve the robustness of the controller, a feedback controller is designed along the planned trajectory. Position to position control of the robot is simulated to demonstrate the effectiveness of the method.

Most current bipedal robots were modeled with an assumption that there is no slip between the stance foot and ground. This paper relaxes that assumption and undertakes a comprehensive study of a compass gait biped with foot slipping. It is found that slips are most likely to happen near impact for a broad range of gaits. Among these gaits, ones with a backward swing foot velocity relative to the ground just before touch down generally require less friction to maintain stable walking than ones with a forward relative foot velocity. Moreover, a larger percentage of gaits with the "swinging backward" foot can tolerate some slipping without falling than those with a swinging forward foot at touch down. Thus, a gait with the swing-backward foot just before touch down should be more robust in the sense of preventing slipping and falling. It is further shown that only one parameter in gait design determines the swing-backward feature, which can help design robust gaits. Models with varying physical parameters such as mass, leg length, and position of center of mass (CoM), are also studied to validate the generality of the results.

This paper presents an approach to show global controllability of a class of underactuated mechanical systems. This class of systems include gymnastic robots and the classic cart-pole system. For this class of systems, we can define a connection point, and design a damping controller u(t) with the Lyapunov method to drive all states to the connection point. Further by exploiting the time reversal symmetry of the system and spline interpolation, a controller u(-t) can be obtained to drive the system from the connection point to any other state. The system is thus shown completely controllable, i.e., there exists an admissible trajectory from any given state to any given final state. Swing-up control designs for a pendubot, cart-pole system and triple pendulum are given as illustrative examples.

This paper presents a study of the relationship between the magnitude of coupling between the actuated and unactuated degrees of freedom for an ankle-actuated biped robot and the robustness and cost of transport of the gait. It extends prior results that considered only the instantaneous coupling at individual poses to analyze the cumulative effects of coupling over entire trajectories. The coupling metric is general in that it can be computed directly from the Lagrangian of the system. By using a two-link biped model, a family of ankle-actuated candidate gaits defined by fourth-order Bézier polynomials is generated, and the feasibility of the gaits is verified in the sense that the unilateral constraint of ground reaction force is satisfied. The correlation between the coupling metric and the maximum magnitude of disturbance that can be rejected is significant. It indicates that robust gaits tend to have small coupling under zero disturbance so that the “reserve” coupling may be utilized to reject the disturbance. Moreover, gaits with smaller cost of transport under zero disturbances have smaller coupling and therefore should be more robust.